131 8
English Pages 135 [132] Year 2023
Romana Ishrat Editor
Biological Networks in Human Health and Disease
Biological Networks in Human Health and Disease
Romana Ishrat Editor
Biological Networks in Human Health and Disease
Editor Romana Ishrat Centre for Interdisciplinary Research in Basic Sciences Jamia Millia Islamia University New Delhi, India
ISBN 978-981-99-4241-1 ISBN 978-981-99-4242-8 (eBook) https://doi.org/10.1007/978-981-99-4242-8 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore
Preface
Advances in high-throughput biotechnologies have led to the generation of huge amounts of biomedical data that open a new research horizon from a reductionist view to a more complex understanding of the biological systems. To describe complex interactions and regulatory mechanisms behind biological systems, biological networks are employed to represent all relevant interactions taking place in biological systems. In networks, molecules (genes, proteins, metabolites) are reduced to a series of nodes connected by edges. Edges represent the pairwise relationships and interactions between two molecules within the same network. Molecular networks have become extremely popular and have been used in every area of biology to model, for example, transcriptional regulation mechanisms and physical protein-protein interactions. Network theory offers a versatile and general toolbox for a framework for investigating biological systems ranging from the molecular to the global scale level. A key factor for the success of network theory in biomedical applications is that many structural network characteristics can be related to the functional properties of the respective biological system. The biological network also leads to a wide range of applications, such as pathways related to a disease that can unveil how the disease acts and provide novel tentative drug targets. In addition, it can also help predict the responses to disease and can be useful for novel drug development and treatments. In this book, the authors discuss various network theoretic and data analytics approaches used to analyze biological networks with respect to available tools, technologies, standards, algorithms, and databases for generating, representing, and analyzing graphical data. New Delhi, India
Romana Ishrat
v
Contents
1
raph Theory in the Biological Networks���������������������������������������������� 1 G Riddhi Jangid, Pallavi Somvanshi, and Gajendra Pratap Singh
2
Biological Networks Analysis�������������������������������������������������������������������� 15 Najma and Anam Farooqui
3
etwork Analysis Based Software Packages, Tools, and Web Servers N to Accelerate Bioinformatics Research���������������������������������������������������� 51 Nikhat Imam, Sadik Bay, Mohd Faizan Siddiqui, and Okan Yildirim
4
etworks Analytics of Heterogeneous Big Data�������������������������������������� 65 N Rafat Ali and Nida Jamil Khan
5
etwork Medicine: Methods and Applications�������������������������������������� 75 N Aftab Alam, Okan Yildirim, Faizan Siddiqui, Nikhat Imam, and Sadik Bay
6
ole of R in Biological Network Analysis������������������������������������������������ 91 R Mohd Murshad Ahmed and Safia Tazyeen
7
achine Learning in Biological Networks���������������������������������������������� 111 M Shahnawaz Ali
vii
Editor and Contributors
About the Editor Romana Ishrat is an Professor at the Centre for Interdisciplinary Research in Basic Sciences at Jamia Millia Islamia University in New Delhi, India. She has extensive experience in teaching and research in the fields of computer science, machine learning, bioinformatics, and related areas. Her research focuses on developing and utilizing novel statistical and computational methods to address genetics and genomics problems related to complex diseases. Dr. Ishrat has published over 60 research articles in prestigious journals such as Springer, Elsevier, Frontiers, Willey, and Oxford Press in the areas of network biology, transcriptional networks, systems biology, data and text mining.
Contributors Mohd Murshad Ahmed Singapore Institute of Clinical Sciences (SICS), Singapore, Singapore Aftab Alam Centre for Interdisciplinary Research in Basic Sciences, Jamia Millia Islamia, New Delhi, India Rafat Ali Department of Biosciences, Jamia Millia Islamia, New Delhi, India Shahnawaz Ali CGTRM, Kings College London, London, UK Sadik Bay Research Institute for Health Sciences and Technologies (SABITA), Istanbul Medipol University, Istanbul, Turkey Anam Farooqui Centre for Interdisciplinary Research in Basic Sciences, Jamia Millia Islamia, New Delhi, India Nikhat Imam Department of Mathematics, Institute of Computer Science and Information Technology, Magadh University, Bodh Gaya, India Centre for Interdisciplinary Research in Basic Sciences, Jamia Millia Islamia, New Delhi, India Riddhi Jangid School of Computational and Integrative Sciences, Jawaharlal Nehru University, New Delhi, India
ix
x
Editor and Contributors
Nida Jamil Khan Department of Biosciences, Jamia Millia Islamia, New Delhi, India Najma Centre for Interdisciplinary Research in Basic Sciences, Jamia Millia Islamia, New Delhi, India Faizan Siddiqui International Medical Faculty, Osh State University, Osh, Kyrgyzstan Mohd Faizan Siddiqui International Medical Faculty, Osh State University, Osh, Kyrgyzstan Gajendra Pratap Singh School of Computational and Integrative Sciences, Jawaharlal Nehru University, New Delhi, India Pallavi Somvanshi School of Computational and Integrative Sciences, Jawaharlal Nehru University, New Delhi, India Safia Tazyeen Center for Interdisciplinary Research in Basic Sciences, Jamia Millia Islamia, New Delhi, India Okan Yildirim Department of Chemical Biology Otto-Hahn-Strasse, Max Planck Institute of Molecular Physiology, Dortmund, Germany
1
Graph Theory in the Biological Networks Riddhi Jangid, Pallavi Somvanshi, and Gajendra Pratap Singh
Abstract
Graph theory is a mathematical tool widely used to study many different areas today. In this chapter, we demonstrate how the basic graph theory concepts can be applied to study molecular biology and its insights. We first introduced graph preliminaries and then used the theory by representing molecular networks using graphs. The depiction has been made by using examples and figures for a better understanding of a beginner. The chapter gives a good connection and motivation to study biology and mathematics together. Keywords
Biological data · Biological networks · Graph theory · Molecular interactions
1.1 Introduction Graph theory is a branch of the study in discrete mathematics that expresses many different networks and structures in different fields like social network analysis and biological networks. We use interchangeably (very often) terms “graph” and “network” in the chapter. A graph is nothing but a set of points linked with a set of lines. These points in a graph depict the entities that we want to model and the lines in the graph represent the connection between the points in our network. For instance, suppose we want to understand the network topology of DNA then we use the concept of an n-dimensional De Bruijn graph in graph theory and can create a DNA R. Jangid · P. Somvanshi · G. P. Singh (*) School of Computational and Integrative Sciences, Jawaharlal Nehru University, New Delhi, India e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 R. Ishrat (ed.), Biological Networks in Human Health and Disease, https://doi.org/10.1007/978-981-99-4242-8_1
1
2
R. Jangid et al.
graph for any sequence (Jafarzadeh and Iranmanesh 2016). In the next section, we study the basic definitions of graph theory in order to understand how a graph can be studied in the context of Biology.
1.2 Basic Concepts of Graph Theory 1.2.1 Graph A graph consists of a non-empty set of elements (called vertices, points, junctions) that are joined with directed/undirected lines (called edges). The collection of these lines forms an edge set that contains ordered/unordered pairs of vertices. Hence, we define a graph G ≔ (V(G), E(G)) where V(G) and E(G) denotes the set of vertices and edges, respectively. E(G) consisting of only ordered pair of vertices is called a directed graph (digraph) and with that of only unordered pair of vertices is called an undirected Graph (Barnes and Harary 1983; Gupta 2008). Example: In Fig. 1.1, the vertex set is {A, B, C, D, E, F} while the edge set for Fig. 1.1a is {(A, D), (C, D), (A, B), (B, C), (A, E), (E, F), (C, F)} where the order of elements can be altered while in Fig. 1.1b it is strictly in the order {(A, D), (D, C), (A, B), (B, C), (A, E), (E, F), (F, C)}. It is easy to note for the undirected graph that every edge can be written in two different orders unlike in directed graphs. For instance, (C, D) and (D, C) are equivalent in the undirected graph. In biological context, directed graphs are used in the modeling of transcriptional regulatory networks and metabolic networks and in the study of neuronal networks. Directed graph because the nodes representing genes in the transcriptional regulatory networks will have a natural direction associated while modeling the interaction from genes X to Y. In the case of undirected edges, the study of protein–protein interaction networks describes the physical interactions among the organism’s proteome.
Fig. 1.1 (a) Undirected graph and (b) Directed graph with six vertices and seven edges
1 Graph Theory in the Biological Networks
3
1.2.2 Degree of a Vertex It is defined as the total number of edges associated with the vertex in a graph. In directed graphs, it is further classified as the Indegree or Outdegree of a vertex. We say Indegree and Outdegree are the number of edges incoming to the vertex and the number of edges outgoing from the vertex, respectively. In the above example, the degree of every vertex in Fig. 1.1a can be written as: deg(A) = deg (C) = 3 and deg(B) = deg (D) = deg (E) = deg (F) = 2 while in Fig. 1.1b; indeg(A) = 0, indeg(B) = 1, indeg(C) = 3, indeg(D) = 1, indeg(E) = 1, and indeg(F) = 1. Also, outdeg(A) = 3, outdeg(B) = 1, outdeg(C) = 0, outdeg(D) = 1, outdeg(E) = 1 and outdeg(F) = 1.
1.2.3 Representation of a Graph When it comes to representation, we represent any graph with its Adjacency matrix and Incidence matrix. If there is any graph with p vertices and q edges, then the adjacency matrix MA = [a]ij is of order p × p while the incidence matrix IA = [b]ij is of order p × q. We define them as, wherever there exists any edge among the vertex, we place the value 1 otherwise, 0. In the case of an undirected graph, the adjacency matrix is obviously a symmetric matrix unlike in the digraph. For the incidence matrix in a digraph, +1 represents the direction of the edge outgoing from a vertex while −1 depicts the direction of the edge incoming to a vertex. These adjacency matrices also give us information on the degree of vertices in the graph. For the example above, the adjacency matrices and incidence matrices for Fig. 1.1. are given in Figs. 1.2a, b and 1.3a, b. Here, we can also find the degree of respective vertices in an undirected graph by their row sums while for a directed graph, we can find the outdegree and indegree by row sums and column sums, respectively.
Fig. 1.2 (a) Adjacency matrix and (b) Incidence matrix for Fig. 1.1a
R. Jangid et al.
4
Fig. 1.3 (a) Adjacency matrix and (b) Incidence matrix for Fig. 1.1b
Fig. 1.4 (a) An example of a subgraph of G. (b) An example of a spanning subgraph of G
1.2.4 Subgraph A graph G′ ≔ (V′(G′), E′(G′)) is said to be a subgraph of G if and only if the V′(G′) and E′(G′) of G′ are the subsets of the V(G) and E(G) of G, respectively. Further, if the V′(G′) = V(G) but the E′(G′) of G′ is a subset of the E(G) of G, then it is called spanning subgraph (see Fig. 1.4a, b).
1.3 Graph Algorithms Theorem 1 The sum of degrees of all the vertices in any graph G, is even. Proof We know that every edge in a graph G contributes 2 degrees to the graph while 1 degree to every vertex to which it is adjacent. So, we can write:
2 ( q ) = deg ( v1 ) + deg ( v2 ) +…+ deg ( v p ) ,
1 Graph Theory in the Biological Networks
5
where p and q denotes the number of vertices and edges, respectively, in the graph G. Hence, 2(q) is always even. Theorem 2 In any graph, there are always even numbers of vertices of odd degrees. Proof Let us consider a graph G having both degrees of vertices, even and odd. Diving the vertices into two groups, one having even degree vertices and the other one with odd degree vertices such that G1 = deg (v1e) + deg (v2e) + … + deg (vpe) and G2 = deg (v1o) + deg (v2o) + … + deg (vpo). By Theorem 1 we know that the sum of degrees of all the vertices is even, hence G1 + G2 is also even. This further implies that since G1 is already even, G2 must be even. Hence, proof.
1.4 Complex Graph Models Graphs of different types are studied according to the complexity of literature requirements. These different types of graphs are defined as follows.
1.4.1 Bipartite Graph Bipartite graph is that type of non-empty graph in which the set of vertices is partitioned into two disjoint subsets, say 𝐴 and 𝐵 such that every edge joins the vertex from set 𝐴 to set 𝐵 and vice versa. A directed bipartite graph is that type of bipartite graph which is directed also (see Fig. 1.5).
1.4.2 Complete Graph Joining any vertex with every other vertex using an edge in a graph, such a graph is called a complete graph. We denote such graphs using Kp, where p denotes the number of vertices in the graph.
Fig. 1.5 Example of a complete Bipartite graph K3, 3
6
R. Jangid et al.
1.4.3 Weighted Graph What if we assign some non-negative integer values to the edges of the graph? In that case, we will call such a graph a weighted graph in which non-negative integers (called weights) are attached to the edges.
1.4.4 Eulerian Graph In a graph starting from a vertex, if we cover every edge exactly once even after repeating the vertices and end with the same initial vertex, then such type of graph is called Eulerian Graph. In Fig. 1.1, the graphs are not Eulerian. One important result for Eulerian graphs is—Every connected graph is Eulerian whose all the vertices are of even degree.
1.4.5 Hamiltonian Graph If instead of edges we wish to cover every vertex exactly once in any graph, then such type of graph is called a Hamiltonian graph. In this case, all edges may not be included. For example, Fig. 1.1a is a Hamiltonian graph since moving from D→A→B→C→F→E covers all the vertices exactly once.
1.4.6 Regular Graph A graph such that all the vertices in the graph have the same degree k is known as a k-regular graph.
1.4.7 Planar Graph A graph that can be drawn on paper without crossing an edge and without lifting a pen even once is called a planar graph.
1.5 Fundamentals of Network Theory and Its Characteristics We illustrate here the fundamentals of network theory with respect to the biological context.
1 Graph Theory in the Biological Networks
7
1.5.1 Biological Networks In the context of network biology, there are different questions that can be asked of graphs like: 1 . Whether there is any way to produce a metabolite X from A? 2. How long is a particular chain to X from A? 3. Whether all the proteins are connected to others by any path? 4. Which is the most influencing protein in any network? We will look for answers to such questions using the graph theoretical concepts that we discussed above. Showing the existence of a path in a graph/network solves our first query while the second problem is the shortest path problem in graph theory. For the third query, one can look for the identification of connected components in a graph/network. And in the last one, we can solve it using looking for the most connected nodes in the graph/network. In such problems, we study graph algorithms that are applicable and used in many different areas (Jafarzadeh and Iranmanesh 2016; Koutrouli et al. 2020; Mason and Verwoerd 2007; Pavlopoulos et al. 2011) as well. For example, the shortest path problem, traveling salesman problem, looking at connected components, minimum spanning tree problem, centrality measures, and Eulerian and Hamiltonian path problems.
1.5.2 Mathematical Concepts in Relation to Network Biology Except from the basic terminology in graph theory, we study: Density: It is defined as the ratio of edges that exists in a network. Mathematically, q we write the formula for the directed graph DenD = while for the undip ( p − 1) 2q rected graph DenU = . p ( p − 1) Degree distributions: It is a histogram of degrees. It gives us the node fraction in p any network having some degree x. Mathematically, P ( x ) = x where px is the p number of nodes with degree x while p is the nodes in the network. Clustering Coefficients: Also called cliquishness, we determine the degree to which the vertices of a graph form clusters with each other. The clustering coefficient tells us how much of a clique we find in a neighborhood of a vertex. 2y where u denotes the node with degree x and y are the Mathematically, Cu = x ( x − 1) number of edges among x neighbors of u in the graph.
8
R. Jangid et al.
1.6 Application of Graph Models in Biology There are numerous applications of graph models in biology. We illustrate here by showing the examples of biological networks (see Fig. 1.6): • • • • • •
Interaction networks for Protein–Protein Networks for Sequence similarity Networks for Gene regulatory system Network for Signal Transduction Network for Metabolic pathways Network for Gene co-expression
There are numerous types of data that can be presented in the theory of network biology and presenting the data as a network and their network analysis is an integral part of Systems Biology. When it comes to Protein–Protein Interactions, it is an undirected network as we can see in Fig. 1.7. When it comes to gene regulation networks, these are directed ones. Going from one gene to another is shown here by the green lines. In the case of the cell signaling pathway, the depicted path in the network is C→D→F. Metabolic biological pathways can be directional, and hence depicted using directed edges. Consider the network between proteins/genes A, B and metabolites m1, m2 and m3 where the directed edge set depicting metabolism pathways is {(A, m2), (m2, B), (m1, A), (m3, B)}, the concept of Bipartite graphs can be implemented here like metabolites interact with proteins/genes and two different kinds of nodes are used here to depict the interactions. We discuss here one such property of the Protein-protein Interaction network (PPIN). We are basically looking at the study of how a protein interacts with the other proteins in a cell. While studying the interactions, we show here a small world effect.
Fig. 1.6 Depiction of types of biological networks using a graph
1 Graph Theory in the Biological Networks
9
Fig. 1.7 (a) Protein–protein interaction network example of shortest path. (b) Protein–protein interaction network showing the proteins with maximum degrees. (c) Protein–protein interaction network showing the highly interconnected proteins in respective clusters
Looking at the figure in the PPINs example, there are majorly three sub-networks in whole. Another formal word we use here is that we have three clusters in this network. This is an undirected graph that has proteins as its nodes and their interconnections as edges. This shows good connectivity among the proteins. Consider looking for an optimal way to reach from protein A to protein B in the Fig. 1.7a. This is what we call the problem of finding the shortest path from one node to another. Here, the length of the path for moving from A to B is 5. The degree of node A and node B in the network is 1 in both cases which depicts that they both are connected to a single protein directly. We can see clearly from the figure that there is no Eulerian or Hamiltonian path that exists from A to B. Finding out the proteins that are important in the network can be studied using their centrality measures in the graph. This we can see in the next Fig. 1.7b, where we have the proteins in different colors showing high connectivity in the network. And in Fig. 1.7c is the nodes with high transitivity meaning the proteins that are densely connected in the respective clusters.
10
R. Jangid et al.
1.6.1 Petri Net Modeling Approach in Pathway Analysis Petri nets, a formalized modeling language from around the 1960s, have been used to describe and model discrete event dynamic systems (DEDS) (Peterson 1981). These are structurally directed bipartite multigraphs, where we associate some defined set of rules with the graph and in turn help in analyzing DEDS. Bipartite in the sense as we have two disjoint sets of nodes depicted using circles (called places) and rectangles (called transitions) and directed arcs depicting the direction of process flow in the graph (Fig. 1.8). The arcs must be used to connect circles and rectangles but in a manner such that it joins no two circles or rectangles at a time. An edge that is directed from a circle to a rectangle is the input place of that rectangle and vice versa. For any graph, known as the Petri net graph, we assign the tokens denoted by dots inside the circles in the graph. Writing the number of tokens as a vector is known as marking the Petri net. These markings are used for further execution of the Petri net. Theoretically, we define marking as n − vector = (M1, M2, M3, …, Mn), where n is number of places in the Petri net structure and each Mi ∈ {0, 1, 2, 3, …, n}. We use the Petri net modeling approach in order to find the possible pathways in biological networks given to us with certain conditions. Though there are some graph models that also study the pathway, there are some limitations to them. Hence we use the Petri net approach.
1.6.1.1 Petri Net Reachability-Based Analysis Petri net firing rule: Removing w(p, t) tokens from the input place of a transition and depositing w(t, p) tokens into the output place of the transition causes the state change in a Petri net, which we call firing of that transition. Here, w(p, t) denotes the weight of the arc from place to transition, and similarly w(t, p) denotes the weight of the arc from transition to a place. In the example we are discussing, we have w(p, t) = w(t, p) = 1 for all the transitions (see Fig. 1.9) (Murata 1989).
Fig. 1.8 Example of a Petri net with initial marking (2, 1, 0)
1 Graph Theory in the Biological Networks
11
Fig. 1.9 Illustration of firing rule in Petri net depicted in Fig. 1.8
Fig. 1.10 Reachability/ Coverability graph of Petri net in Fig. 1.8
1.6.2 Petri Net Invariant Analysis Another analyzing technique for Petri net models is by creating Incidence matrix and solving them. We will get some equations but these equations have limited solvability because of restrictions like the solution should be a non-negative integer. Hence, a Petri net model must be pure in order to use this analyzing technique. We write an incidence matrix where the rows and columns are the places and transitions, respectively. And based upon these matrices, we mathematically solve by creating state equations and studying the Invariants. In Fig. 1.10, we have S1, S2, S3, S4 as the reachable markings from the initial marking S0. From Fig. 1.9, we can say that S0 = (2, 1, 0) and S1, S2, S3, S4 equals (1, 1, 1), (1, 0, 1), (0, 1, 2), (0, 0, 2), respectively. The red color of S3 and S4 in Fig. 1.10 depicts the state of deadlock in the Petri net while the red color of transitions in Fig. 1.9 depicts that the transitions are enabled in the respective state of the Petri net. In order to study the pathways in biological processes, we study with the reachability-based analysis (Jha et al. 2022; Mansoori et al. 2020) as well as the
12
R. Jangid et al.
Fig. 1.11 Example of Hamiltonian path approach for DNA sequencing
Incidence matrix and state equations (Singh and Gupta 2019; Singh et al. 2022; Singh et al. 2020; Singh et al. 2021).
1.7 Case Study As a case study, we look for a DNA sequencing example. In DNA sequencing, we shear the DNA into millions of small fragments. Now we read 500–700 nucleotides at a time from the small fragments (Sanger method). Our challenge is to assemble these short fragments (called reads) into a single genomic sequence (called “superstring”). Let us suppose that we are given data on some pieces of the Genome. If we wish to generate a DNA sequence from those genome pieces, then we have two graph theoretic approaches to solve this problem: 1. Hamiltonian path approach—Taking the given pieces as nodes (with length k) and these nodes are connected by an arc if the k-1 rightmost nucleotide of first vertex overlaps with the k-1 nucleotide of the second one. 2. Eulerian path approach—Here, we take suffix/prefix as a node. Each oligonucleotide becomes an arc in which its initial endpoint is k-1 rightmost nucleotide of the arc and its terminal endpoint is k-1 leftmost nucleotide. For example (Fig. 1.11), we have the given pieces of genome—M = GTG, GCG, GCA, ATG, TGG, TGC, GGC, CGT, CAA, AAT. Hamiltonian path approach is used to generate a DNA sequence from these genome pieces and, we have the graph shown in the figure. Using this approach, we have the following graph model where we follow the path in which every node is covered exactly once. The resulting Hamiltonian path will be—ATG(v0)-TGG(v1)-GGC(v2)-GCG(v3)-CGT(v4)-GTG(v5)TGC(v6)-GCA(v7)-CAA(v8)-AAT(v10). Hence, we get the following sequence of Genome—ATGGCGTGCA.
1 Graph Theory in the Biological Networks
13
Fig. 1.12 Example of Eulerian path approach for DNA sequencing
But this approach of getting the sequence is time consuming when we have these pieces in millions or billions. So we use the second approach which is comparatively easier. In the Eulerian path approach, we have the graph as shown in Fig. 1.12. We follow the path as shown in the figure where every edge is covered exactly once and we got this as follows: ATG(v0)-TGG(v1)-GGC(v2)-GCG(v3)-CGT(v4)GTG(v5)-TGC(v6)-GCA(v7)-CAA(v8)-AAT(v10). Hence, we get the following sequence of Genome—ATGGCGTGCA. Clearly, both approaches give us the same solution but the second approach is less time consuming and easier to follow. So, in this way graph theory is used in DNA sequencing. Many other concepts can be studied in different other case studies as well.
1.8 Conclusion We have discussed in the chapter only the basic mathematical concepts in graph theory that are used in biology. One can use the theory in many different ways according to their needs. In order to study complex systems in biology, there are many extensions available in the literature that we might have missed but this note definitely covers all the basics that are difficult to find at a place. Acknowledgments The third author, Dr. Gajendra Pratap Singh is thankful to the Department of Science and Technology(DST)-Science and Engineering Research Board (SERB) Project (PID:M T R/2021/000378) and to the Indian Council of Medical Research (ICMR) project (PAC/SC&IS/ GPS/ICMR/1511); and the authors are expressing their deep gratitude to anonymous reviewers and editors for their valuable suggestions and comments.
References Barnes JA, Harary F (1983) Graph theory in network analysis. Soc Networks 5(2):235–244 Gupta SB (2008) Discrete mathematics and structures. Laxmi Publications, Ltd Jafarzadeh, N., & Iranmanesh, A. (2016). Application of graph theory to biological problems. Studia Universitatis Babes-Bolyai, Chemia, 61(1)
14
R. Jangid et al.
Jha M, Singh M, Singh GP (2022) Modeling of second-line drug behavior in the treatment of tuberculosis using Petri net. Int J Syst Assur Eng Manag 13(2):810–819 Koutrouli M, Karatzas E, Paez-Espino D, Pavlopoulos GA (2020) A guide to conquer the biological network era using graph theory. Front Bioeng Biotechnol 8:34 Mansoori F, Rahgozar M, Kavousi K (2020) A pathway analysis approach using Petri net. IEEE J Biomed Health Inform 25(3):874–880 Mason O, Verwoerd M (2007) Graph theory and networks in biology. IET Syst Biol 1(2):89–119 Murata T (1989) Petri nets: properties, analysis and applications. Proc IEEE 77(4):541–580 Pavlopoulos GA, Secrier M, Moschopoulos CN, Soldatos TG, Kossida S, Aerts J et al (2011) Using graph theory to analyze biological networks. BioData mining 4(1):1–27 Peterson JL (1981) Petri net theory and the modeling of systems. Prentice Hall PTR Singh GP, Gupta A (2019) A Petri net analysis to study the effects of diabetes on cardiovascular diseases. In: In 2019 6th international conference on computing for sustainable global development (INDIACom). IEEE, pp 481–488 Singh GP, Jha M, Singh M (2020) Modeling the mechanism pathways of first line drug in tuberculosis using Petri nets. Int J Syst Assur Eng Manag 11(2):313–324 Singh GP, Jha M, Singh M (2021) Petri net modeling of clinical diagnosis path in tuberculosis. In: Advances in interdisciplinary research in engineering and business management. Springer, Singapore, pp 401–412 Singh GP, Jangid R, Singh M (2022) Petri net modeling as an aid in bioprocess designing. In: Microbial products. CRC Press, pp 455–462
2
Biological Networks Analysis Najma and Anam Farooqui
Abstract
Networks are widely recognized as a popular method for representing complex biological processes by capturing the interconnected relationships between different biological components using binary interactions or connections. This chapter discusses many forms of biological networks and network models that are important for understanding complicated networks. In addition, we describe different network measures that are quantifiable description of biological networks. We also discussed different methods for detection of community. We briefly mention about hubs and formation of rich-club, system-level organization in a hierarchical network, detection of network modules and motifs, and biomarkers. Lastly, we examine several databases that contain biological networks and explore how network modules are utilized in understanding the dynamics of diseases. We anticipate that a wide range of readers, from experts to newcomers, will benefit from this chapter and be influenced to advance the field. Keywords
Biological networks · Graph theory · Protein–protein interaction networks (PPI) · Node · Edge · Genetic interaction networks · Degree distribution · Closeness centrality · Eigenvector centrality · Neighborhood connectivity · Betweenness centrality rich club · Transcriptional regulatory networks · Network motifs · Hierarchical network · Cell signaling networks · Hamiltonian energy · Clustering coefficient · Clustering methods · Biomarker · Metabolic networks · Network model · Scale-free network · Biological network databases · Disease dynamics Najma · A. Farooqui (*) Centre for Interdisciplinary Research in Basic Sciences, Jamia Millia Islamia, New Delhi, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 R. Ishrat (ed.), Biological Networks in Human Health and Disease, https://doi.org/10.1007/978-981-99-4242-8_2
15
16
Najma and A. Farooqui
2.1 Overview of Biological Networks: Network Construction For a layperson, genomes, in general, are thought of as components that are purely defined by their linear DNA sequences. However, the reality is far more complex. The DNA (Deoxyribonucleic Acid) serves as the primary unit of heredity in all types of organisms. The DNA helix folds itself into several layers of higher-order structures in a hierarchical fashion that forms a chromosome (Woodcock 2006). The DNA folded in this fashion eventually gives it a compact structure enabling it to get accommodated in the limiting space of the cell’s nucleus. Besides this sophisticated arrangement of the genetic component itself, the cellular factors that are responsible for reading, copying, and maintaining the genome too are arranged in a complex and compartmentalized fashion within the cell’s nucleus. The cellular organization of the genomes and the cellular factors give an architectural environment for them to function. Determining how these particular molecules contribute to various cellular processes is one of the main problems in current cell biology. Identification of these molecules and their interactions is critical for gaining complete knowledge of the complicated machinery within live cells. Exploring the answers to these questions may lead to insights into genome biology and how it works (Fig. 2.1). We, as human beings, are surrounded by a network of networks. Our body functions through complex networks of networks working at different levels. Most biological functions arise from complex interactions from intricate interactions involving multiple cellular elements like nucleic acids, proteins, and small Fig. 2.1 DNA makes up genes and is coiled within chromosomes inside the nucleus of a cell. Credit: NIGMS.
Chromosome Nucleus
Cell
DNA
Gene
2 Biological Networks Analysis
17
molecules. It is exceedingly uncommon for a biological function to be attributed solely to a single molecule. These intercellular webs of interactions include PPI, metabolic networks, transcription-regulatory networks, and signaling networks. The technologies like PROTEIN CHIPS or YEAST TWO-HYBRID determine these molecular interactions. Another key technique that is commonly used for analyzing complex systems in the physical, social, information, technology, and biological sciences is network theory. The application of network theory in biology is known as Network Biology. The development of network biology has revealed that there are universal laws that control cellular networks. This might provide a fresh conceptual framework for comprehending disease pathology in the twenty-first century. The biological systems are depicted in this as standard graphs with nodes and edges. The graphs’ nodes or vertices stand in for various entities or agents, and the edges or links signify a connection between two nodes. In a more realistic and complicated framework, these edges may be heterogeneous, unweighted/weighted, and directional/ unidirectional. Biological networks play a crucial role in examining and understanding the interconnected relationships among various components (interactome) inside a biological system. These interactomes are the collection of direct or indirect molecular linkages of the biological system. Thus, overall, a biological network signifies the cell’s information processing system via the molecular wiring or molecular network (Kaviani and Sohn 2021).
2.1.1 Structure of Complex Networks and Notations Because of the data explosion generated by the omics era of biological research, it was important to shift away from a single gene/protein perspective and develop more systematic data analysis methodologies. System biology seeks to explore biological processes on a systems level, not just as isolated components but also as interacting systems and their evolving characteristics. System biology, which uses graph theory approaches to describe and analyze biological systems, is connected to network biology. The study of graphs, which are mathematical constructions used to represent pairwise interactions between things, is known as graph theory. In this context, a graph is a collection of vertices or nodes interconnected by lines or edges. In actuality, it is a group of conceptual ideas and techniques for visualizing and analyzing networks. Nodes represent different entities (e.g., genes or proteins) and edges convey information about how the nodes are linked (Muzio et al. 2021). Biological network analysis was driven by the methods and concepts of social network analysis, and the application of graph theory to the social sciences also contributed to its development. From the molecular to the ecosystem level, every biological entity interacts with other biological entities, giving us the opportunity to describe biology using a variety of networks such as ecological, neurological, metabolic, or molecular interaction networks. Complex interactions among biomolecules are usually described using network models when studying biological systems.
18
Najma and A. Farooqui
These networks are known as biological networks and they represent biological systems mathematically (Pavlopoulos et al. 2011). For example, networks within cells typically exhibit the following characteristics: 1 They tend to be disassortative. 2 They possess structural and dynamic robustness. 3 Their degree distribution power law. 4 They exhibit a modular organization. 5 The average length of the shortest path between any two nodes is quite short, indicating a small-world feature.
2.2 Biological Networks and Types 2.2.1 Biological Networks The idea of biological networks is based on the observation that multiple networks operate in living cells to carry out and regulate every element of cellular life, these elements are made up of a variety of components ranging from basic biomolecules to the entire organism, and these networks function within a highly organized framework. All biological activities are carefully and tightly controlled at the cellular level (Grigorov 2005). The foundation of life processes for the entire structural range of living matter, including biomolecules like proteins and nucleic acids, cellular organelles like cytoskeleton and mitochondria, tissues, whole cells, and organs, is made up of biological networks, which are organized, deterministic systems. Finally, the entire organism makes up this structural spectrum. These networks collaborate to carry out certain physiological activities inside a particular cellular compartment. They describe how various biological processes, such as genetic regulation, cellular signaling, and metabolism are structured. These networks are frequently used to comprehend and analyze biological processes at the system level, find fascinating modules or sub-networks, and identify potentially important proteins based on their network characteristics. These networks are representations of biological systems and the relationships that exist between them. They provide insights into complicated biological systems, revealing information about them. They integrate biological omics data with biological interactome data to disclose information inside these systems (e.g., gene–gene associations and protein–protein interactions). The networks involved in transcription, protein–protein interaction, and metabolism are the most important for managing biological systems. These networks are studied using a combination of statistical methodologies, graph theory methods, mathematical models, and visualization tools. The study of biological networks is now an important part of systems and computational biology. Because such analysis provides a common language for describing relationships within complex systems, it has become more crucial in gaining a better understanding of physiological functions.
2 Biological Networks Analysis
19
Many efforts have been made to study and analyze the topology and structure of cellular networks, as well as their relationship to cellular function and organization. These networks are often modular, have a small world property with minimal average path-lengths, and use scale-free topologies with power law degree distributions; this makes them resistant to adversity. As a result, networks can enhance our knowledge of biological systems and their flow. Among other aspects, networks are used to analyze gene lists from high- throughput biology as well as data from post-genome-wide association studies. There are various approaches to using network information to enhance our knowledge of biological systems. One example is the development of new organizational assumptions based on network topology information. Using massive network data to test and confirm or refute current hypotheses is a complimentary strategy for a very long time, theorists have speculated about the connection between the evolution of genes and the networks that they produce. With the availability of large-scale quantitative data on the topology of molecular networks, it is now possible to pose explicit queries about the role of network structure in the evolutionary process and how evolution influences network structure. Several recent researches have demonstrated the effectiveness of this innovative approach to biology (Yu et al. 2013).
2.2.2 Types of Biological Networks 2.2.2.1 Different Types of Biological Networks Are Described below 1. Protein–protein interaction networks. 2. Metabolic networks. 3. Gene/transcriptional regulatory networks. 4. Genetic interaction networks. 5. Cell signaling networks. Protein–Protein Interaction Networks One of system biology’s main objectives is to better understand protein–protein interactions. PPINs, or protein–protein interaction networks, are mathematical depictions of the physical interactions that take place between proteins in a cell. Understanding protein–protein interactions (PPIs) is critical for understanding both normal and pathological cell physiology because PPIs are required for understanding every process in a cell. It is also essential in drug development. PPIs are a type of molecular interaction data that is widely used. These interactions give both an experimental foundation for understanding cell modular architecture and important information for predicting the biological function of specific proteins. Many of the cell’s most critical molecular activities, such as DNA replication, are carried out by complex molecular machines made up of many protein components that are organized by their protein–protein interactions. It is possible to represent a collection of pairwise interactions among a collection of proteins in a natural way by using a graph with proteins as its nodes and pairwise interactions as its edges. The collection of all interactions (entire set of PPINs that exist in a biological system) between the proteins of an organism is usually called the interactome (Jordán et al. 2012).
20
Najma and A. Farooqui
PPIs can be detected by: in vitro methods (like NMR spectroscopy, protein fragment complementation, phage display, tandem affinity purification, protein arrays, X-ray crystallography, and co-immunoprecipitation, affinity chromatography) and in vivo methods (like yeast two-hybrid (Y2H, Y3H) and synthetic lethality) based approaches). These experimental resources have been used to create extensive PPI networks. But on the other hand, the volume of PPI data is making laboratory validation difficult. Therefore, computational analysis of PPI networks using in silico methods (like in silico 2 hybrid structure-based approaches, gene expression, gene fusion, sequence-based approaches, chromosome proximity, phylogenetic tree, and mirror tree), is turning into a crucial tool for understanding the functions of yet undiscovered proteins. Protein–protein interaction (PPI) is currently one of the important areas of study for the advancement of contemporary system biology (Srinivasa Rao et al. 2014). Metabolic Networks Metabolic networks are built by collecting and linking biochemical data with genetic sequences. The mass flow in basic chemical pathways that generate vital components like amino acids, carbohydrates, and lipids, as well as the energy required by biological reactions, are usually the focus of the metabolic network. As a result, these networks frequently include information on both proteins and metabolites. The metabolic network is made up of all chemical processes that utilize catalytic proteins to facilitate the metabolism of small molecules (metabolites). A metabolic pathway is a set of processes that transform one or more educts into one or more products. Metabolic pathways are involved in the storage and release of energy. Metabolic networks are made up of very similar building components. The vertices in this network indicate the educts and products, while the connecting reaction is represented by an edge that will be tagged with the catalyzing enzyme and perhaps supplemented with cofactors. This representation solely depicts a reaction’s connection (topology). The reaction kinetics, or the concentration gradients of the relevant molecular partners in terms of their starting concentrations, is required for modeling (Haggart et al. 2011). Metabolic network databases include databases that focus on a single organism like HinCyc110 (for H. influenzae), EcoCyc109 (for E. coli), and PseudoCyc111 (for Pseudomonas aeruginosa), as well as databases that include a wide range of organisms like KEGG, MPW, and MetaCyc. These databases contain information and graphics that depict such pathways and their connections. Computers are increasingly being used to process photos. These databases are large enough to compare metabolic pathway topologies in various animals (Succoio et al. 2022). Gene Regulatory Networks Gene regulatory networks, or GRNs, are essential for regulating how genes are expressed. These are collections of regulatory links between transcription factors (TFs) and TF-binding sites, or between genes and their regulators. Cis- and
2 Biological Networks Analysis
21
trans-regulatory elements are the two primary regulatory types. Trans-regulatory elements can regulate away from the genes from which they were transcribed, whereas cis-regulatory elements are located close to the structural region of the gene they regulate. Moreover, GRNs are directional as TFs regulate their targets, dynamic, i.e., changing across different conditions, and can be visualized as bipartite graphs, which are crucial for understanding the underlying mechanisms of disease pathogenesis. GRNs are composed of “nodes,” which represent the genes and their regulators, joined together by “edges,” which represent physical and regulatory interactions (Ertaylan et al. 2014). GRNs are bipartite because there are two types of nodes: Genes and TFs. The overall topology of GRNs has been analyzed in many systems. To create the most comprehensive and accurate GRN models possible, it is ideal to incorporate physical and regulatory links. The presence of TF and gene hubs indicates that GRNs are not random structures. GRN modules are highly interconnected. There are various databases present that hosts information about gene regulation, commonly used repositories are the KEGG, GTRD, and TRRUST. A comprehensive understanding of GRNs is a major challenge in the field of systems biology. Reconstruction of regulatory networks is made possible by recent developments in high-throughput approaches, which offer a wealth of binding data from techniques like ChIP-Seq, miRNA-Seq, and ATAC-Seq along with expression data from RNA-Seq. Reconstruction of the gene regulatory interaction to form GRNs is one of the key tasks in systems biology. Gene expression data are frequently used in the creation of a GRN. For repairing GRNs in a true cellular context, a number of computational techniques and models have been developed to date. Nevertheless, each of them uses a unique set of presumptions and techniques to create unique blueprints. Several statistical and mathematical tools can be used to visualize and study it. For GRN visualization, Cytoscape is a widely used and simple-to-use application that has shown to be quite helpful (MacNeil and Walhout 2011). Genetic Interaction Networks A genetic interaction network is a collection of genes that are connected by edges and have functional interactions with one another. It is believed that these genes either physically interact with one another through the gene products they produce, such as proteins, or that one gene affects the activity of another gene. The phrase “genetic interaction” refers to a collection of functional connections between genes that cooperate to carry out a certain activity and are frequently physically linked to one another to create a more complicated structure. Understanding biological activities requires information about interacting proteins, which may be easily attained by examining networks of these connections. These interactions are crucial to the majority of biological processes. Genetic interactions require combination of two or more genes to generate an unexpected phenotype. These interactions are further categorized into two categories i) Negative genetic interaction) and ii) Positive genetic interactions. Negative genetic interactions are caused when two mutations, none of them lethal individually, combine to cause cell death. Positive genetic interactions occur when a lethal
22
Najma and A. Farooqui
variant of one gene is suppressed by the variant of another gene or their combined effect is less severe than expected. These types of interactions are essential in understanding pathways and regulation in model organisms, functional relationships between genes, also undiscovered genetics associated with complex human diseases (Boucher and Jenna 2013). A resource for forecasting gene and pathway function is a global genetic interaction network, which shows the functional arrangement of a cell. This network highlights the ubiquity of genetic connections and how they can amplify the abnormalities brought on by a single mutation. They typically show pairs of genes active in parallel pathways or various molecular machinery (Costanzo et al. 2016). Genetic interaction profile of a gene is made up of its unique set of positive and negative genetic interactions. Positive interactions may provide links associated with problems in cellular proteostasis and cell cycle progression or insights into general mechanisms of genetic suppression. Negative genetic interactions between functionally related genes, mapped key bioprocesses, and recognized pleiotropic genes can be inferred using alternative functional information. Genetic interaction profiles provide a quantifiable measure of functional similarity, and similar networks created by correlating large-scale genetic interaction profiles organize genes into clusters that highlight biological processes. When genetic interaction profiling networks are coupled with other types of interactome networks, predictive models of biological processes can be created, resulting in potentially powerful models which can form a potentially powerful model by using different datasets (Wiredja and Bebek 2017). Cell Signaling Networks A crucial regulatory system that is vital for all life activities in living creatures is cell signaling. Cell signaling networks control and direct cellular functions, intercellular interactions, and reactions to the environment. The various signaling pathways that signaling networks use to carry out their functions each represent an ordered series of reactions that are elicited and started by signal molecules, primarily proteins that turn on receptor proteins, and that lead to biochemical or biophysical modifications in the pathway. A primary mechanism to control the number of intracellular components occurs at different levels, i.e., the post-transcriptional mRNA processing level (alternative splicing) than on next level which includes post-translational modifications (such as phosphorylation, acetylation, methylation, and so on). Signaling networks can be visualized as directed graphs with edges pointing in the direction of signal propagation. There are one or more starting nodes in a signaling network that represent the binding of the initial signal(s) to receptor(s) and one or more output nodes that indicate the cellular responses to the signal(s). Along with these nodes, there are several intermediary nodes that involve ions, enzymes, genes, secondary messengers, kinases, proteins, metabolites, and other chemicals in signal transmission. The edges of a signaling network indicate many interactions between signaling components such as transcription, protein phosphorylation, protein binding, enzymatic catalysis, complex formation, and regulation (Vieira and Vera- Licona 2019).
2 Biological Networks Analysis
23
In living organisms, there are numerous cellular signaling networks, both extracellular and intracellular. Some important cellular signaling networks are the PI 3-Kinase/Akt Signaling pathway, MAPK/ERK pathway, Wnt signaling pathways, Apoptosis signaling pathways, cAMP-dependent signaling pathway, Retinoic acid signaling pathway, Calcium signaling pathways, Delta-Notch signaling pathway, Hedgehog signaling pathway, Insulin signaling pathways, VEGF mediated signaling pathway, Oscillatory rhythm network, and many more. At each step of the cellular signaling process, feedbacks are possible which forms a signal transduction network. Cellular signaling network reconstruction includes genome annotation, expression arrays, cell physiology characterizations, biochemical experimentation, and other such data sources. Data for reconstructing networks can be found in renowned pathway databases like (Reactome, KEGG), specific repositories such as the NetPath, MiST, or Human-gpDB are also present (Papin et al. 2005). Modeling cellular signaling networks is quite challenging as they involve the interactions of components from different levels such as transcriptome, metabolome, and proteome. Mathematical modeling of cellular networks enables us to figure out the basic design of cell signaling networks from a system-level perspective and how transmission of information affects the network (Albert and Wang 2009). Large-scale networks can be structurally analyzed in their totality since it does not require a comprehensive understanding of the parameters that have been obtained via thorough experimentation. Complex networks being ubiquitous have gained extensive attention from the scientific community. Complex systems’ internal states can help us gain a general understanding of their biological, technical, and social surroundings. The key component of network biology is the behavior of a complex network. The characteristics of complex networks can be understood through network models. Three main models have been studied that have direct control of biological networks.
2.3 Network Topological Properties From living cells to the Internet, complex systems work in synchronization with its components through pairwise interactions. As mentioned earlier, these components can be described as a set of interconnected nodes. Here, each edge represents the interactions between two nodes. Altogether, these nodes and edges form a network or a graph. Such networks/graphs are studied and conceptualized through Graph Theory. Depending on the type of interactions, networks can be classified as directed or undirected. A directed network’s connections have a specific direction between its nodes, such as when information flow is regulated from a transcription factor to a gene. However, the interactions in undirected networks have no assigned direction. In protein interaction networks, for example, a link represents a mutually binding relationship (Fig. 2.2).
24
Najma and A. Farooqui
Fig. 2.2 Types of graphs (a) Graph representing edges and nodes (b) Directed graph (c) Undirected graph (d) Weighted graph
The structural properties of complex networks are understood by examining their topological parameters. The fundamental network measures that provide insights into the significant behaviors of the network are Degree distribution, Clustering coefficient, Neighborhood connectivity, Eigenvector centrality, Closeness centrality, and Betweenness centrality. These measures help us gain a deeper understanding of the network’s important properties and how its nodes are connected. Degree distribution: In a network, the degree k refers to a measure of centrality that indicates the number of connections a node has with other nodes. In a network represented by a graph G = (N, E), where N and E represent the number of nodes and edges, respectively, the probability of the degree distribution (P(k)) of the network can be defined as the proportion of nodes with a particular degree relative to the total size of the network:
25
2 Biological Networks Analysis
nk (2.1) N Here, nk depicts the count of nodes that have a degree of k, while N denotes the total number of nodes in the network. The probability distribution P(k) provides insight into the significance of hubs or modules within the network. It follows a power law pattern, where P(k) is approximately proportional to the inverse of k raised to the power γ. The value of γ determines the level of importance attributed to hubs or modules in the network, which can vary depending on whether the network exhibits characteristics of scale-free or hierarchical structures. Neighborhood connectivity: A node’s connectivity is determined by the number of its neighbors. The neighborhood connectivity of a node “n” is calculated by finding the average connectivity of all of its neighboring nodes. In the network, the neighborhood connectivity (CN(k)) can be expressed as follows: P (k) =
q (2.2) C N ( k ) = ∑ qP q k In the network’s context, the conditional probability (P(q/k)) represents the probability that a link originating from a node with connectivity k will connect to a node with connectivity q. The presence of a positive power relationship in the neighborhood connectivity (CN(k)) can serve as an indicator of assortativity within the network’s topology. Clustering coefficient: The clustering coefficient is a topological parameter that represents the extent of interconnectivity and strength of connections between a node and its neighboring nodes in a network. It is determined by calculating the ratio of the number of edges that exist between the node’s nearest neighborhood edges (ei) to the total possible number of edges for a node of degree ki. In an undirected network, the clustering coefficient (C(ki)) of the “ith” node can be calculated using the below formula:
C ( ki ) =
2e i k i ( k i − 1)
(2.3)
Betweenness centrality: Betweenness centrality (CB) is a measure of a node’s importance in controlling the flow of information within the network. It reflects the extent to which a node can control other nodes by acting as a bridge between them. To compute the betweenness centrality of a node “v”, we consider the number of geodesic paths between every pair of nodes “i” and “j” that pass through “v”, represented by dij(v), and the total number of geodesic paths between “i” and “j”, represented by dij. The formula for calculating the betweenness centrality (CB(v)) of a node “v” is:
CB ( v ) =
∑
i , j, i ≠ j≠ k
d ij ( v ) d ij
(2.4)
Closeness centrality: Closeness centrality (CC) quantifies the efficiency of information dissemination from a node to other nodes that it can reach within the
26
Najma and A. Farooqui
network. It is defined as the reciprocal of the average geodesic distance between the node and all other connected nodes in the network. In other words, the closeness centrality of a node “i” is calculated as the inverse of the average shortest path distance between node “i” and its neighboring nodes. The formula to determine the closeness centrality (CC) of a node “i” is as follows: CC ( i ) =
n ∑ j d ij
(2.5)
where dij represents the geodesic path length from nodes i to j, and n is the total number of vertices in the graph reachable from node i. Here, dij denotes the geodesic path length from node i to j, while n represents the total number of vertices in the graph that are accessible from the node i. Eigenvector centrality: The eigenvector centrality (CE) of a node “i” in a network is directly related to the collective centrality of its neighboring nodes. The formula for calculating the eigenvector centrality (CE) of node “i” is as follows: 1 (2.6) ∑ vj λ j= nn (i ) Here, the term “nn(i)” refers to the nearest neighbors of node “i” within the network. The eigenvalue (λ) and eigenvector (vi) of the node are determined by the equation Avi = λvi, where “A” represents the adjacency matrix of the network or graph. The principal eigenvector, associated with the highest eigenvalue (λmax), is considered to have a positive eigenvector centrality score. It can be used as an indicator of spreading power a node in the network (Farooqui et al. 2018). CE ( i ) =
2.4 Detection of Network Module and Motifs Network motifs refer to small, interconnected sub-graphs that are frequently observed in networks. These motifs represent recurring patterns within the network structure. However, identifying network motifs requires significant computational effort as they need to be tested for similarity multiple times (Patra and Mohapatra 2020). The various tools and algorithms used in the process of finding network motifs are briefly described below: NeMoFinder: Mesoscale network motifs are discovered using this algorithm. Network motif detection (NetMODE): It is a network motif detection software package to improve runtime efficiency. Grochow and Kellis: It is a motif-centric algorithm, where frequency is counted on the basis of a particular isomorphic class. Kavosh: It is a network-centric algorithm to improve runtime efficiency. Elhesha–Kahveci: It is a motif-centric algorithm for finding disjoint network motifs in a target network. MODA: It is a motif-centric algorithm based on a pattern growth methodology.
27
2 Biological Networks Analysis
CoMoFinder: An algorithm to accurately and efficiently identify composite network motifs in genome-scale coregulatory networks. MODET:—It is an motif-centric algorithm based on a static ET. Accelerated motif (Acc-Motif) and QuateXelero: These are network motif discovery algorithms. MDET: It is a Fast and scalable network motif discovery algorithm for exploring higher-order network organizations.
2.5 Rich Club and Community Finding Algorithm 2.5.1 Rich-Club Finding Algorithm In a network, the nodes that have a large endowment to the overall topological organization are called hubs. The most generally used centrality in identifying hubs is degree centrality, and the other centralities (like closeness, eigenvector, and betweenness) also give useful detail on each node. Other measures like vulnerability, are used to identify the vitality of a node in a network. Since there is no single direct way to identify hubs, each measure is very much correlated to one another. Therefore, it is always good to use them together to rank the important nodes and identify the candidate hubs (Rubinov and Sporns 2010). Due to their central role, hubs perform many important roles in the network. These hubs are accounted for effective communication, as well as the inclusion of information, among different nodes or modules in a network. Many real-world complex networks, such as social and World Wide Web, are observed to have dispersed topologies consisting of an organization of functional modules. In such networks, the importance and roles of hubs are very illuminative, and these hubs have a tendency of interconnecting among themselves further forming a highly powerful group of specialized hubs, known as a rich club. The formation of rich clubs increases the robustness, efficient communication, propagation of the signal, and integrability of a complex network (van den Heuvel and Sporns 2011). In a modular network, the hubs can be further classified using the participation coefficient measure (Pi) defined as (Guimerà et al. 2007): kij (2.7) Pi = 1 − ∑ N k i Here N stands for the total number of modules, ki denotes the hub node’s degree, and kij denotes the number of links the hub node possesses with other nodes within a specific module (j). After identifying the degree range k for hub nodes, the existence of rich clubs is studied by measuring the rich-club coefficient (k) over the degree range (Zhou and Mondragon 2004). j =1
2
28
Najma and A. Farooqui
2.5.1.1 System-Level Organization in a Hierarchical Network System-level organization of fundamental functional components of a hierarchical network at different levels is maintained to perform important specific tasks at those various levels in the self-organized fashion of the components (System-Level Organization in a Hierarchical Network n.d.) and the emergence of important regulators at a local and global level referred to as hubs (Albert and Barabási 2002) are some of the fundamental features of most of the natural and artificial networks. This hierarchical organization of the network extends from how cells function, brain organization, and how proteins cross-talk at the molecular level (Stelzl et al. 2005), the evolution of species during the prebiotic era (Jain and Krishna 2001) to inter and intracellular talks in tissue network. Traditionally, biology has been based on the central idea that life processes are hierarchically organized and indicates that it is this structure that controls the system’s dynamics. Surprisingly, we lack an objective manner to assess how real this hierarchical organization is, even if we are given different levels and their interactions in the hierarchy (Pennisi 2005). Among the topological characteristics of a network one is functional organization of it via various fundamental functional units known as network motifs (Milo et al. 2002) and their roles in building up the network organization/reorganization that led to the network complex in nature. Motifs in a network may be of many types, and every network motif performs a well-defined function within the network (Alon 2019) and most motifs in the network overlap to process information among them. Clustering of motifs (similar types or different) by overlapping structural and functional modules of various topologies, clustering modules form super-modules that cross-talk among them to organize the whole network. Cancer networks are observed to have self- similar organizational characteristics like rich-club formation, modular structure, hubs, etc. This small-world structure generates nonlinear dynamical behavior and the rich-club formation supports the flexible integration of the individual modules. Complex network theory is a useful tool to study the organizational structure of complex systems like cancer. The exploration of self-organizing properties can be done by observing fractal growth mechanisms (Song et al. 2006).
2.5.2 Community Finding Algorithm Clustering: detection of community: Networks that are created from the real world have distinctive properties (topological) which are dissimilar from random networks given by Pál Erdős and Alfréd Rényi in 2002. They generally show dependence on degree for clustering coefficients with biased diversity in degree, which is absent in random networks. These networks (real-world) show edges heterogeneously on a large scale in distribution, with nodes forming dense sub-regions of closely connected groups resulting in a modular structure in the network calling it a community. As seen clearly from fractal studies and scaling behaviors, many of the real-world complex networks have shown to be organized as hierarchical with self- similar sub-units (sub-communities) present in larger communities (Ravasz and Barabási 2003).
29
2 Biological Networks Analysis
Finding communities is, however, not a simple problem. Primary goal of this is to extract useful information to efficiently perceive the functional organization in the system that it represents, however, it may indirectly help in identifying the abnormalities in the system and also a possible target. A community can be perceived as the existence of densely connected nodes into a consistent group with relatively fewer connections with the outside group. Therefore, it is required to apply efficient algorithmic methods, which is a great challenge considering the huge amount of computational complexity (Fortunato 2010). This computational complexity estimates the maximal number of computational steps required to perform an assigned task that accounts for the worst-case scenario(s) and is affected by the way a network is stored in the working memory of the computer. In network-based algorithms, networks are usually stored in either adjacency list format or adjacency matrix (adjacency list occupies less space). The long history of dividing a network into modules starts from the work of Stuart A. Rice in the 1930s who grouped people on voting patterns similarities, by studying their working relationships using personal interviews with Weiss and Jacobson. In 1962, H. Simon discussed the importance of maintaining hierarchy, self-organization, evolution in complex systems, and many more. The first algorithm called Kernighan–Lin algorithm was developed in the 1970s for dividing resources to have effective parallel computing. However, being specific, it was a graph partitioning algorithm. While the idea behind finding communities is to identify the natural fault lines between the communities without having previous knowledge. Therefore, it is natural to come up with so many ways for finding a community based on their definition of a community. The techniques for finding community can be classified into four major types: (1) Agglomerative that relies on some similarity measures: such as Jaccard similarity, cosine similarity, correlation coefficient, Katz similarity, or Euclidean distance; (2) Identification of intermodular edges: methods based on fluid-flow, betweenness centrality measures, current-flow analogies, etc.; (3) Quality function optimization of partitions, like likelihood-based measures, modularity, and Hamiltonian in Potts model; and (4) statistical inference methods such as block modeling (Bickel and Chen 2009). The most commonly used quality function is the modularity that compares an observed edge density within a partition. Modularity Q can be measured by using: ki k j 1 ∑ Aij − 2m ij 2m
(2.8) δ ( Ci ,C j ) Here m stands for a total number of edges in the community, k represents degrees, Aij denotes adjacency matrix of size i × j, and the δ function yields 1 if nodes i and j are in the same community. Some well-known community finding algorithms are discussed below. Hierarchical clustering method: With the rise of social network analysis, this clustering method focuses on the hierarchical decomposition of a network aggregating nodes based on predefined topological similarity measures. There are varieties of clustering (hierarchical) algorithms that emerge from the pliability in the choice Q=
30
Najma and A. Farooqui
of the “similarity measure” as well as having different rules for assigning “similarity score” to a collected mass of nodes. Therefore, a hierarchical clustering method begins with a commonly used measure, i.e., cosine similarity σji: (2.9) σ = n / k k i j ij ij where kj and ki are the respective degree of nodes j and i. The algorithm then generates an n × n similarity matrix, from this similarity matrix the nodes bearing the highest similarity values are agglomerated to form group(s) of size = 2. The next step of agglomeration involves a choice from three methods: (1) single-linkage, (2) complete-linkage, and (3) average-linkage clustering. Let us suppose in the general form we have two groups of size vi and vj so there will be vivj possible pairs such that one node falls into group vi and another node into group vj. From these pairs, highest similarity pair is considered in the case of single-linkage clustering, while the least similar pair is considered for complete-linkage clustering. Finally, the mean similarity from all the pairs is considered as an index in the case of average-linkage clustering. Finally, we obtain a complete hierarchical picture of the network, starting from single nodes and successively grouping them into clusters in the hierarchical order (Newman 2012). This method is still widely used, as it tends to have clusters of high-degree nodes leaving behind low-degree nodes. Betweenness-based method: In 2002, M. Girvan and M. E. J. Newman suggested the most popular method for recognizing communities by performing topological measures of the network (Girvan and Newman 2002). They used the concept of betweenness centrality and formulated the concept of edge betweenness which is described as the number of geodesic paths that run through an edge. It further considers the emergence of compact modules expected to have a higher edge betweenness. The algorithm first calculates the edge betweenness and search for the highest score. Then remove the corresponding highest-scoring edge and on each edge removal, the edge betweenness scores are again calculated, this process iterates, then, in the end, the network starts to split into parts until the nodes are separated. Finally, the network is decomposed hierarchically and can be represented by a dendrogram which is left at the bottom representing the individual nodes. One disadvantage of this algorithm is its high computational complexity which means it is a slow algorithm. The method was further extended by using modularity maximization to identify the best partition. Another vital variation of this model was proposed by Radicchi et al. in which the intermodular edges are identified in dense modules that can have the formation of short loops as compared to the edges between modules. Other related models that use the “fluid-flow” can successfully identify intermodular edges and also improve the speed of the algorithm (Wu and Huberman 2004). Optimization method: Optimization method is based on heuristics algorithms involving the calculation of an approximate solution by assigning a quality function. Modularity, Hamiltonian in Potts model, E/I ratio, likelihood base measures, etc., are some classes of the quality function used in optimizing problems. From these, modularity is mostly used because of its innate ability to interpret communities based on a null model. There is a variety of algorithms based on approximation
2 Biological Networks Analysis
31
techniques from various disciplines such as physics, mathematics, biology, and computational science. Some common methods are simulated annealing (Guimerà et al. 2004), greedy algorithm, spectral optimization techniques, etc. For instance, simulated annealing is a global optimizing strategy based on the concept of slow cooling of solids, in physics. It uses a probabilistic strategy to provide a good approximation for a global optimum solution (Medus et al. 2005). It starts by arbitrarily assigning some communities to the network. Then, randomly moves node i into another community with a condition for the new community must have at least one edge already connected to node i. If the modularity Q increases the move is accepted, otherwise a probability eβ_Q is assigned to the move. The algorithm stops when no further improvement on modularity is possible, and finally, the state reached is the best-approximated state. The algorithm has been reported to give good results (Danon et al. 2005); however, it is slow. While greedy algorithm on the other hand starts considering each node as a partition. Then in every followed step, a single edge is added until all edges are added. Then, the optimized configuration is chosen by selecting the maximum modularity state. In addition to it, Schuetz and Caflisch (Schuetz and Caflisch 2008) suggested instead of allowing one community pair, more pairs should be allowed to avoid large accumulation into large communities. Other important community detection techniques include Modularity optimization and methods like Infomap algorithm a non-modularity-based, statistical inference and distance-based clustering algorithm, and other statistical inference methods.
2.6 Biomarker Discovery Biomarker was defined as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention” by a working group of National Institutes of Health Biomarkers in 1998. According to WHO, definition of biomarkers includes “almost any measurement reflecting an interaction between a biological system and a potential hazard, which may be chemical, physical, or biological? The measured response may be functional and physiological, biochemical at the cellular level, or a molecular interaction.” Biomarkers are a measurement tool for assessing the health status of an individual measured from outside of the patient, i.e., they provide reliable and accurate indications of the state of their health (Strimbu and Tavel 2010) Currently, it is used as a chemical that is administered into an organism to check organ function or other health-related characteristics whose detection signals a certain disease’s development, progression, and treatment effectiveness. For instance, an infection may be indicated by the presence of an antibody. More specifically, biomarkers can include certain cells, chemicals, genes, gene products, enzymes, or hormones. They can also include anything from blood pressure pulse and to more comprehensive laboratory testing of blood, urine, tissues, and other body fluids. Biomarkers can identify both complex organ processes and distinctive alterations in biological systems.
32
Najma and A. Farooqui
Biomarkers are used in the pharmaceutical industry to determine a drug’s efficacy and safety, which can reduce development costs and lead to additional therapeutic targets. By using biomarkers, scientists can increase their precision in addition to providing more accurate diagnoses and recommending more efficient therapies. The need for novel therapeutic approaches, such as regenerative medicine, the availability of potent new “omics” technologies, the rise of novel and unproven targets in the pharmaceutical industry, and the chance to use improved and well- validated biomarker assays have all played a role in the dramatic rise in the discovery of biomarkers over the last 10 years. In order to encourage the early removal of unpromising compounds and accelerate the release of breakthrough medicines and technology, the tool Support Industry has prioritized the safety and mode of action, i.e., potency-related indicators over disease-related biomarkers. Surrogate markers are required to replace inaccurate clinical end objectives, particularly in fields of newly evolving technologies such as regenerative medicine. Precisely, biomarkers are required to determine which treatments are most suited to each individual’s needs (Krzyszczyk et al. 2018). The use of new and complicated technologies is improving biomarker discovery and development, raising the possibility that additional clinical uses of biomarkers will be applied to improve illness diagnosis, prognosis, and monitoring. Biomarker development is a multi-step process that includes basic research, validation, and clinical application.
2.7 Identification of Key Regulators All hubs within a network play crucial roles in regulating its functions, but the most powerful and impactful genes are those that govern the network’s operations at both the overall and motif-specific levels. These genes, referred to as the “Key Regulators,” were identified through gene tracing techniques. By employing Newman and Girvan’s community detection or clustering method, gene tracing was performed within numerous communities or sub-communities, extending up to the motif level (Tazyeen et al. 2022). By employing tracing techniques, it is possible to identify most influential and significant genes in the network that control the network. Other Methods like Rich-club finding algorithm (Sect. 2.5.1), Community finding algorithm (Sect. 2.5.2), and Biomarker Discovery (Sect. 2.6) which we have discussed earlier in this chapter are also used for identification of Key regulators.
2.8 Statistical Properties and Models of Biological Network 2.8.1 Random Network Model The Random Network Model was first developed by Pál Erdős and Alfréd Rényi. Since the large random networks follow the Poisson degree distribution, it is often called the Poisson random network model (Wu et al. 2017). Generally, a network
2 Biological Networks Analysis
33
consists of only nodes and links. However, to reproduce the complexity of a real system, the links of the network must be placed appropriately. In the case of Random Networks, the links are randomly placed between the nodes (Fig. 2.3a). Two different models define the Random Network Model. The first is G(N, M) model in which a random network G is connected randomly with N vertices and M edges. Another model is G(N, p) in which instead of specifying M edges, the nodes are specified with probability p in the random network. Thus, the size of the preferred random network is produced both by specifying the number of edges, or the probability of observing the links between the edges. The random networks can be analyzed by investigative multiple graphs, where the number of nodes/vertices remains identical, but their links are randomly changed (Fig. 2.3a). The Random Network Model has led to several significant findings for network structure. The Random Network follows Poisson degree distribution and is completely connected for fairly small values of average degree. This signifies that for a random network each node does not need to be connected to too many other nodes. Another significant feature of a random network is that the connected random network is slightly compact even for large networks.
Fig. 2.3 (a) Random network. (b) Scale-free network. (c) Hierarchical network
34
Najma and A. Farooqui
2.8.2 Scale-Free Network Model Despite the success of the random network model, there remains uncertainty whether real networks such as biological networks or social networks are truly random. This led to the discovery of the scale-free property which describes the structure and dynamics of complex systems (Barabási 2009). Since then, researchers have revealed scale-free structures in a wide range of systems. In the case of the Barabási–Albert model of a scale-free network, the network is governed by high- degree nodes which means nodes that are highly connected (Fig. 2.3b). These highly connected nodes are called hubs (Fornito et al. 2016). For the reason that the real- world networks are scale-free, some degree of nodes with degree k follows a power law k−α, where α > 1. Scale-free networks are robust against accidental failures but susceptible to coordinated attacks. The links in scale-free networks are established based on two different mechanisms, i.e., growth and preferential attachment. A new node prefers to get attached to a node that already has many connections. This will ultimately result in a network that is dominated by a few highly connected nodes called hubs.
2.8.3 Hierarchical Network Model Several real-life networks like protein–protein interaction networks, metabolic networks, or some social networks exhibit scale-free properties along with high clustering. This led to the evolution of hierarchical network models which incorporate the scale-free topology and high clustering into one single model (Ravasz and Barabási 2003). For generating the hierarchical network models, it is expected that clusters associate in an iterative style for the co-occurrence of local clustering, modularity, and scale-free topology in various real systems (Fig. 2.3c). We can also say that the hierarchical network model shares its foremost property of having a greater number of hubs in the network with the scale-free model family. However, unlike other alike models (Barabási–Albert, Watts–Strogatz), in hierarchical models, the nodes with a greater number of connections are likely to have a lesser clustering coefficient. In the case of the Barabási-Albert model, with the increase in the number of nodes the average clustering coefficient decreases. On the contrary, in a hierarchical network, there is no exact pattern and relationship between the network size and its average clustering coefficient.
2.9 Biological Network Databases The analysis and modelling of biological networks, as well as their investigation, are crucial tasks in modern life sciences. The majority of biological networks are still far from full, and because of the complexity of the relationships and the unique characteristics of the data, they are frequently challenging to understand (Zhang and Itan 2019). Therefore, a major problem for bioinformatics is the creation of
2 Biological Networks Analysis
35
sufficient storage and querying technologies. Initially, flat files and relational databases were used to construct storage systems. Despite their simplicity, both types have limitations in terms of access times and querying capability. As a result, a lot of databases have come onto the scene recently (Guzzi and Roy 2020). Here, we will talk about categories of biological network databases as well as some helpful online resources that have biological network information.
2.9.1 Biological General Repository for Interaction Datasets It is a free database that consists manually curated proteins from several species, like yeast, mouse, fly, worm, and human. (Oughtred et al. 2021a). In Biological General Repository for Interaction Dataset (BioGRID), curated interactions can be used to create complicated networks capable of speeding up the discovery of new biomedical treatments, especially for conditions affecting human health and disease. Its data is derived solely from primary experimental data found in the biomedical literature, and it includes both narrowly focused low-throughput trials and large high-throughput datasets. It also tracks how proteins change post-translationally and how proteins or genes interact with bioactive small molecules, including many well-known medications. All annotations are incorporated using an integrated network visualization tool that allows users to generate network graphs of protein, chemical, and genetic linkages (Oughtred et al. 2021b).
2.9.2 The Database of Interacting Proteins This database contains a list of protein interactions that have been verified through experimentation. It integrates data from several sources to create a single, reliable list of protein–protein interactions. The information on protein–protein interaction networks was retrieved from the most trustworthy, core subset of the Database of Interacting Proteins (DIP) data and used in computational ways to both manually and automatically curate the data housed in the DIP database (DIP:Home n.d.). DIP contains PPI from various organisms that have been experimentally confirmed. It is implemented as a relational database. Each DIP entry comprises generic protein information (e.g., gene name and cellular location), as well as cross-references to other databases and information about experimental methods and specific experiments. Each interaction is issued a unique code. DIP interactions must be detailed in peer-reviewed papers, and the entry process is manual. A web interface can be used to create queries in both interactive and batch modes. A user can download a subset of DIP in many formats in batch mode.
36
Najma and A. Farooqui
2.9.3 Biomolecular Interaction Network Database Biomolecular Interaction Network Database (BIND) comprises protein interactions that have been annotated with molecular function information gathered from the literature. It is based on three types of data records: pathways, molecular complexes, and interactions. The database allows various types of searches, including those that use identifiers from other biological databases or those that focus on certain fields like literature data, molecule structure, and gene data, including functions. With the use of a BIND interaction viewer, the extracted data may be seen. Graphs are used to describe networks, and the nodes in the graphs, which represent molecules, are labeled with various pieces of ontological data (Bader et al. 2001). A web-based system is readily available for searching, examining, and submitting records. Individual contributions, interaction data from the PDB, and various large-scale interaction and complex mapping studies using mass spectrometry, yeast two hybrid, genetic interactions, and phage display have all been added to BIND. The graphical analysis tool, which helps connect functional domains to protein interactions, allows users to visualize the domain composition of proteins in interaction and complex data. In addition, a tool for grouping interaction networks has been developed to aid in focusing on crucial areas. (Bader et al. 2003).
2.9.4 IntAct IntAct is a free database and toolbox for storing, presenting, and analyzing protein interactions. It contains not only protein interactions data but also DNA and molecular interactions data (Orchard et al. 2014). All interactions are either submitted directly by the user or derived from literature curation. The web interface of IntAct provides users with textual and visual depictions of protein interactions, allowing exploration of interaction networks in relation to the Gene Ontology (GO) annotations of the involved proteins. Additionally, a web service is available to facilitate computational retrieval of interaction networks in XML format. Currently, IntAct contains approximately 2200 binary and complex interactions, which have been meticulously curated in collaboration with the Swiss-Prot team and extensively annotated using controlled vocabularies to maintain data consistency (Hermjakob et al. 2004). It establishes a system for semantically consistent annotation by utilizing regulated vocabularies and ontologies. Researchers can submit PSI-MI interactions to the database curators via email.
2.9.5 Online Predicted Human Interaction Database It contains predicted interactions between human proteins. It mixes PPI obtained from databases and books with hypotheses generated by other organisms. Online Predicted Human Interaction Database (OPHID) graph visualization tool allows
2 Biological Networks Analysis
37
users to view the results of queries made using one or more protein IDs. For academic use, the database is available without charge (Brown and Jurisica 2005).
2.9.6 Search Tool for the Retrieval of Interacting Genes/ Proteins (STRING) It is a repository that intends to collect biochemical relationships between proteins, proteins and DNA, and DNA and DNA. On the website, you can access the database by entering a protein identification or the main sequence of a protein. The STRING database’s goal is to gather, score, and integrate all publicly available sources of protein–protein interaction data, as well as to supplement them with computational predictions. Its ultimate goal is to create a comprehensive and objective global network that includes both direct (physical) and indirect (functional) interactions. The most crucial feature is the ability to upload complete genome-wide datasets as input, which allows users to visualize subsets as interaction networks and do gene set enrichment analysis on the entire input (Szklarczyk et al. 2019).
2.9.7 Molecular Interaction It is a curated database of PPIs for multiple model organisms. Molecular Interaction’s (MINT’s) objective is to systematically collect and organize information regarding molecular interactions by retrieving experimental details from peer-reviewed journal publications. Over time, the number of curated physical interactions in the database has increased to approximately 95,000. This comprehensive dataset is publicly accessible online through web-based interfaces and an FTP server, enabling users to explore the data interactively or in batch mode. MINT also contains HomoMINT, a database that focuses on interactions between human proteins discovered through analysis of orthologous proteins in model species (Chatr-aryamontri et al. 2007).
2.9.8 Regulatory Network Repository This resource focuses on five main types of post-transcriptional regulatory connections in humans and mice. It contains in-depth details on numerous combinations of synergistic organizational interactions between TFs, miRNAs, and genes. The flexible architecture of Regulatory Network Repository’s (RegNetwork’s) database allows for future expansions to encompass gene regulatory networks of other organisms. It encompasses a comprehensive compilation of experimentally observed or predicted regulatory interactions at the transcriptional and post-transcriptional levels. Utilizing RegNetwork, researchers can delve into context-specific investigations of transcriptional and post-transcriptional regulatory interactions by leveraging domain-specific experimental data (Liu et al. 2015).
38
Najma and A. Farooqui
2.9.9 Transcriptional Regulatory Relationships Unraveled by Sentence-Based Text Mining It is a repository of human and mouse transcriptional regulatory networks. It contains 8444 TF-target regulatory interactions from 800 human TFs and 6552 interactions from 828 mouse TFs. It is now the most comprehensive publicly accessible database of human transcription factor (TF)–target interactions. It enriches TF–target pairs significantly, especially in highly ranked interactions inferred from high- throughput data. This shows that Transcriptional Regulatory Relationships Unraveled by Sentence-Based Text Mining (TRRUST) can be used to reconstruct human transcriptional regulatory networks (TRNs) computationally (Han et al. 2018).
2.9.10 miRTarBase This database collects verified interactions between microRNA and its targets. Six hundred fifty seven miRNAs and 2297 target genes from 17 species, including human, mouse, chicken, sheep, and others, are stored in this database. It compiled a database of around 3500 MTIs by manually examining relevant literature and painstakingly data mining the text to filter research articles linked to functional investigations of miRNAs. The MTIs collected in the miRTarBase can also provide a considerable number of positive samples for the development of computational algorithms capable of finding miRNA–target interactions. This database also makes use of gene ontology and KEGG pathway enrichment annotation to investigate the functionality of target genes involved in human MTIs (Hsu et al. 2011).
2.9.11 BioCyc Pathway/Genome Databases (PGDBs) Thousands of sequenced organisms’ metabolic pathways and genome information are available in this database. These are created by the same software that anticipates the metabolic pathways of completely sequenced organisms and identifies the operons and genes that provide the necessary enzymes. Additionally, information from other bioinformatics databases is incorporated, including Gene Ontology details and the protein features from UniProt. On its website, a variety of software applications are available for database searching and visualization, omics data processing, comparative genomics, and comparative pathway research (Karp et al. 2017).
2.9.12 MetaCyc It is a reference database that includes metabolic processes and enzymes from different domains of life. It lists the primary and secondary metabolic pathways as well
2 Biological Networks Analysis
39
as the metabolites, operations, enzymes, and genes associated with them. It is the most extensive curated compilation of metabolic pathways, with 2749 pathways drawn from over 60,000 publications. Its content is carefully selected and evidence- based, resulting in an encyclopedic reference tool for metabolism. It is also used to produce several hundred organism-specific Pathway/Genome Databases (PGDBs) that can be assessed from BioCyc.org. MetaCyc aims to create a database of the entire world of metabolism by conserving a representative sample of each experimentally elucidated clarified pathway (Caspi et al. 2020).
2.9.13 ENZYME The ENZYME database is a library of information about the nomenclature of enzymes. It has been recently developed into a crucial tool for the creation of metabolic datasets. In the most recent version, there are details on 3705 enzymes. It is mostly based on recommendations made by the Nomenclature Committee of the International Union of Biochemistry and Molecular Biology (IUBMB), and it describes every type of recorded enzyme for which an EC (Enzyme Commission) number has been assigned (Bairoch 2000).
2.9.14 Reactome It is a freely available and reviewed database having knowledge about pathways in living organisms. It shows how different molecules, like DNA, proteins, and small molecules, work together in biological processes such as traditional intermediate metabolism, signaling, innate and adaptive immunity, and apoptosis. Its objective is to offer bioinformatics tools that will facilitate fundamental research, genomic analysis, modelling, systems biology, and education by visualizing, interpreting, and analyzing pathway knowledge. With a focus on producing precise and reliable answers for genome-wide datasets with interactive reaction times, it offers a variety of pathway analysis tools. Pathway analysis techniques are commonly used to analyze Omics data generated by high-throughput technology (Fabregat et al. 2017).
2.9.15 KaPPA-View4 It is a database that stores metabolic pathways and enables the visual representation of gene-to-gene and metabolite-to-metabolite relationships as curves on a metabolic pathway map or a combination of up to four maps. This illustration is helpful in uncovering new roles for transcription factors that control the genes in a metabolic pathway. Its website http://kpv.kazusa.or.jp/kpv4-kegg/ provides access to KEGG pathway maps and maps produced from their gene classifications (Sakurai et al. 2011).
40
Najma and A. Farooqui
2.9.16 Netpath Pathway NetPath is a database of carefully selected human signaling pathways. NetPath provides extensive maps of several immunological signaling pathways, including over 2800 examples of transcriptionally regulated genes and roughly 1600 reactions retrieved from the literature all of which are connected to more than 5500 research publications. It is the result of collaboration between Johns Hopkins University’s Pandey Lab and the Institute of Bioinformatics. NetPath’s all pathways can be downloaded free of cost in SBML version 2.1 and BioPAX level 3.0, PSI-MI version 2.5 (Kandasamy et al. 2010).
2.9.17 TRANSFAC The database TRANSFAC offers details on the genomic binding locations and DNA-binding preferences of eukaryotic transcription factors. It has a library of positional weight matrices that includes a unique collection of DNA-binding models that can be used to conduct a complete examination of genomic sequences for probable transcription factor binding sites (TFBSs). It can be used as a transcriptional regulatory encyclopedia or as a tool to identify possible TFBSs (Matys et al. 2003).
2.9.18 Human Protein Reference Database Human Protein Reference Database (HPRD) is a unified platform for visually representing and integrating information about domain architecture, post-translational modifications, interaction networks, and disease associations for each protein in the human proteome. HPRD serves as a consolidated platform that visually presents and integrates various data about post-translational modifications, domain architecture, interaction networks, and disease associations for every protein present in the human proteome. The data within this database is meticulously curated from scientific literature by proficient biologists who carefully read, comprehend, and analyze the published information. HPRD was constructed using an object-oriented database in Zope, an open-source web application server that provides flexible querying capabilities and enables dynamic display of data (Peri et al. 2004).
2.9.19 DisGeNET DisGeNET is among the most complete databases of its sort currently available, containing over 38, 0000 linkages between more than 16,000 genes and 13,000 variants and disorders of genes implicated in human disorders. It incorporates data from GWAS catalogs, animal models, scholarly literature, and expert-curated sources. Its
2 Biological Networks Analysis
41
data is routinely tagged with community-driven ontologies and controlled vocabularies. It is a platform that can be utilized in a number of research tasks, including investigating the molecular causes of particular human diseases and their comorbidities, analyzing disease gene characteristics, developing hypotheses about drug therapeutic effects and side effects, confirming computationally predicted disease genes, and evaluating the effectiveness of text mining techniques. The data can be accessed using Cytoscape, a web interface application.
2.9.20 Drug Bank Drug Bank is an extensively annotated database that integrates comprehensive drug data with detailed information about drug targets and drug actions. It has been extensively employed in various applications such as identifying in silico drug targets, conducting drug docking or screening experiments, predicting drug metabolism and interactions, facilitating drug design, as well as supporting pharmaceutical education in a comprehensive manner (Wishart et al. 2008).
2.9.21 The Molecular Signatures Database (MSigDB) It is one of the gene set databases for gene set enrichment analysis. It is among the most popular and extensive repositories of gene sets primarily used for gene set enrichment analysis. Initially focused on metabolic disorders and cancer, it has significantly grown over time and now encompasses over 10,000 sets of genes. This database provides a range of gene expression signatures, including empirically obtained signatures and signatures describing pathways and ontologies from other curated resources.
2.9.22 Kyoto Encyclopedia of Genes and Genomes Kyoto Encyclopedia of Genes and Genomes (KEGG) serves as a valuable resource for conducting systematic analyses of gene activities through the exploration of gene and molecular networks. The primary component of KEGG is the PATHWAY database, which provides graphical representations of biochemical pathways, including a wide range of metabolic pathways and selected regulatory processes. Another essential aspect of KEGG is the ortholog group tables, which present information on orthologous and paralogous gene groups across multiple organisms. These tables contribute to the expression of pathway information within the KEGG database (Kanehisa and Goto 2000). KEGG handles the GENES database, which includes gene catalogs for organisms with complete genomes and some with incomplete genomes. Apart from collecting data, KEGG also provides computational tools. These tools help in
42
Najma and A. Farooqui
reconstructing biochemical pathways from entire genome sequences and predicting gene regulatory networks.
2.9.23 NCBI Gene Expression Omnibus Established in 2000, the Gene Expression Omnibus (GEO) database is a freely accessible repository that serves as a global resource for gene expression studies. Its primary purpose is to preserve and distribute high-throughput gene expression data sets and other functional genomics data. Over time, GEO has expanded to accept high-throughput data related to various applications, such as genome–protein interactions, genome methylation, chromatin structure, and keeping pace with advancing technologies. The database contains a vast collection of tens of thousands of research data sets and offers web-based tools and techniques for users to discover data relevant to their specific interests. Furthermore, users can leverage these tools to visualize and analyze the data within the database (Clough and Barrett 2016).
2.9.24 EBI Array Express Array Express serves as a publicly accessible repository dedicated to microarray- based gene expression data. It stores meticulously annotated raw and normalized data, which can be submitted online in a standardized format or directly from local databases or LIMS (Laboratory Information Management Systems). Reviewers and writers are granted password-protected access to prepublication data. Access to the stored data can be obtained through accession numbers or by employing various search criteria like species, author, and array platform. Additionally, a subset of curated data deposited in the Array Express data warehouse offers the capability to query experiments based on gene and sample attributes. For further analysis and visualization of the data, Array Express provides an integrated data analysis tool called Expression Profiler (Rocca-Serra et al. 2003).
2.10 Role of Network Modules in Disease Dynamics In molecular networks, the concept of modularity is widely accepted. Module-based approaches have a number of advantages, including improved disease classification and robustness in the discovery of dysregulated pathways. Module-centric techniques are particularly promising in their investigation because it is thought that complex disorders are caused by a variety of genetic changes altering a common element of the biological system. How can disease-related modules and sub- networks be found? Sub-networks impacted by a specific disease can be differentiated from the broader network by combining the interaction data with extra information available on the disease state.
2 Biological Networks Analysis
43
To detect modified network modules and elucidate the correlation between phenotypic and genotypic data (Cho et al. 2012), scientists have employed a combination of molecular phenotypic data, such as gene expression patterns observed in diseased samples, and genotypic information such as single nucleotide polymorphisms (SNPs) and copy number alterations. By mapping the genes exhibiting alterations in diseases onto a protein–protein interaction (PPI) network and subsequently identifying network modules enriched with these altered genes, it becomes possible to uncover dysregulated pathways. This approach operates under the hypothesis that complex diseases arise from a collection of mutations that, while varying considerably among patients, tend to dysregulate shared pathways. On the other hand, molecular-level alterations like changes in gene expression are directly associated to organismal-level phenotypes like diseases. Therefore, the modules that are enriched with differentially expressed genes are taken into account by a different group of methodologies and it is possible to think of molecular pathways as informational channels. For example, the activation of the EGFR (epidermal growth factor receptor) by its receptor triggers the activation of various downstream signaling proteins, initiating multiple signal transduction cascades such as the MAPK, Akt, and JNK pathways, ultimately leading to cell proliferation. Consequently, the third category of methods focuses on predicting the molecules and modules that govern the transfer of such specific information (Wee and Wang 2017). What advantages do phenotypic and genotypic variations in disease have in relation to their molecular interactions? Firstly, through a groundbreaking approach, Ideker et al. (Ideker et al. 2002) integrated yeast protein–protein and protein–DNA interactions with gene expression changes resulting from perturbations in the yeast galactose utilization pathway. They successfully identified active sub-networks, which encompassed interconnected genes with significant differential expression. Interestingly, these sub-networks included common transcription factors that exhibited modest changes in gene expression but played a crucial role in connecting other dysregulated genes (Hughes and de Boer 2013). Secondly, employing a module- based method enhances statistical power, enabling the identification of perturbed modules even when individual gene perturbations are statistically insignificant. This is particularly relevant in genetic disorders such as autism and schizophrenia, where rare germline alterations pose challenges in distinguishing them from background noise. Recent studies have shown that many of the altered genes are part of highly interconnected protein networks. Consequently, a network-based approach becomes more effective in identifying the causal genes in such scenarios. Thirdly, recognized network modules provide a deeper understanding of the underlying biological mechanisms of diseases, leading to more precise indicators for disease diagnosis and treatment. This enables the identification of specific targets and strategies for intervention. This holistic perspective offered by network modules contributes to improved accuracy and effectiveness in disease management.
44
Najma and A. Farooqui
2.11 Case Study Based on Inference and Analysis of Network Related to Disease Dynamics When used to analyze breast cancer metastasis, Chuang et al. found that dysregulated network modules provide predictions that are more reliable and precise than single gene-based classifications (Chuang et al. 2007). This research established the efficacy of using network modules to classify diseases. Later on, various modifications and changes were proposed to their work. Like (Lee et al. 2008), curated pathways and a subset of genes having characteristics that may be used to distinguish between different disease phenotypes were used by Lee et al. and Dao et al. (Wu et al. 2013) created new network-based methods for classifying cancer subtypes by locating highly connected sub-networks using randomized algorithms and for ideal marker identification methods like bottom-up enumeration and set cover were also presented. Kim et al. found gene modules using a module cover technique to identify disease heterogeneity in Rembrandt brain cancer data and TCGA ovarian cancer samples (Kim et al. 2013). These selected modules were then integrated with the results of an independently proposed classification scheme and this led to the discovery of disease type characterization on the basis of module combinations. Disease homology can also be discussed using network modules. Overlaps of dysregulated network modules help to explain why some complicated diseases have similar phenotypic characteristics. A PathBlast model was used by Suthram et al. (Sharan et al. 2005) to locate dense sub-networks. By evaluating the expression patterns of various diseases in the modules, analysis of disease homology was performed. It was shown that some dysregulated modules are shared by various diseases, this reveals why some medications are effective for a variety of ailments. The PPI network approaches can also be used to identify key regulatory genes of the network that regulates the activities and signal flow of the complete network. In one of our studies, we analyzed the Turner Syndrome (TS) regulatory network constructed from manually curated genes that were involved in TS from published literatures. Protein–protein interactions, functions, TS networks, and orthologs were all combined to accomplish this. It was found that the TS network shows hierarchical features, signifying system-level organization involving the occurrence of modules/communities interrelated in a certain fashion. Two key regulators namely, KDM6A and BDNF were identified. These key regulators serve as the backbone of the network and are deeply rooted. Any network activities are regulated by these genes and could be possible target genes for resisting the syndrome. Because the network is hierarchical, removing KDM6A and BDNF does not cause the network to break down; rather, the network reorganizes itself to accommodate the alteration. We combined the network-based study with phylogenetic study to find a few essential and conserved interactions (interologs). KDM6A-WDR5, KDM6A-ASH2L, and WDR5-ASH2L are three significant interologs (evolutionarily conserved protein–protein interactions) that we identified, constituting a motif (Farooqui et al. 2018).
2 Biological Networks Analysis
45
In another study, we examined the Turner Syndrome network created from the microarray expression data of TS. We again identified the key regulators of TS through microarray data of TS by combining functions, protein interactions, orthologs, disease networks, and its correlation with two common comorbidities, Recurrent Miscarriage (RM) and Diabetes Mellitus Type 2 (T2DM). Some important signature genes of the above comorbidities were identified. It was found that the TS network shows hierarchical features, signifying system-level organization involving occurrence of modules/communities interrelated in a certain fashion. POU2F1, LCP2, CCL22, ENAM, PTPN22, CXCL5, S1PR4, FAM20A, and EFNB3 were identified as important regulators (Farooqui et al. 2021). These key regulators act as the backbone of the network and are deeply rooted (Ideker et al. 2002). Any network activities are regulated by these genes and could be possible target genes for resisting the syndrome. The study of complicated diseases can be benefited by the effective tools provided by network biology. The concept behind network-based techniques is that rather than looking at individual genes, dysregulated modules can provide a more comprehensive understanding of complicated disorders.
2.12 Conclusion Great efforts have been made over the last two decades to extract the dependence and interplay between structure and function in biological networks because they are highly relevant to biological processes. To summarize, biological networks provide a conceptual and intuitive framework for studying, modelling, characterizing, and comprehending complicated interactions between various components of a biological system. Network biology studies the “interactome,” which is a collection of direct and indirect molecular connections in biological systems. Many areas of biomedical science have benefited from network biology. This simple but powerful concept enables us to extract the essence of gene–protein interactions, predict drug interactions, study disease comorbidity, and discover important associations. Of course, identifying an association is only the first step toward identifying a mechanism, but it is frequently a critical step. Because of significant advancements in data capture, computational tools, and network models, network biology has emerged as the first and most crucial step in bioinformatics, providing an approach for understanding the structure–function relationship in biological systems. Recent developments in this subject demonstrate that it can be used to infer biological organization, function, and evolution.
References Albert R, Barabási A-L (2002) Statistical mechanics of complex networks. Rev Mod Phys 74(1):47–97. https://doi.org/10.1103/RevModPhys.74.47 Albert R, Wang RS (2009) Discrete dynamic modeling of cellular signaling networks. Methods Enzymol 467:281–306
46
Najma and A. Farooqui
Alon U (2019) An introduction to systems biology: design principles of biological circuits, 2nd edn. CRC Press, Boca Raton Bader GD, Donaldson I, Wolting C, Ouellette BFF, Pawson T, Hogue CWV (2001) BIND—the biomolecular interaction network database. Nucleic Acids Res 29(1):242–245 Bader GD, Betel D, Hogue CWV (2003) BIND: the biomolecular interaction network Db. Nucleic Acids Res 31(1):248–250. https://bio.tools/bind. Accessed 06 Jul 2022 Bairoch A (2000) The enzyme database in 2000. Nucleic Acids Res 28(1):304–305. https://doi. org/10.1093/nar/28.1.304 Barabási A-L (2009) Scale-free networks: a decade and beyond. Science 325(5939):412–413. https://doi.org/10.1126/science.1173299 Bickel PJ, Chen A (2009) A nonparametric view of network models and Newman–Girvan and other modularities. Proc Natl Acad Sci 106(50):21068–21073. https://doi.org/10.1073/ pnas.0907096106 Boucher B, Jenna S (2013) Genetic interaction networks: better understand to better predict. Front Genet 4:290. https://doi.org/10.3389/fgene.2013.00290 Brown KR, Jurisica I (2005) Online predicted human interaction database. Bioinformatics 21(9):2076–2082. https://doi.org/10.1093/bioinformatics/bti273 Caspi R et al (2020) The MetaCyc database of metabolic pathways and enzymes - a 2019 update. Nucleic Acids Res 48(D1):D445–D453. https://doi.org/10.1093/nar/gkz862 Chatr-aryamontri A et al (2007) MINT: the molecular INTeraction database. Nucleic Acids Res 35(Database):D572–D574. https://doi.org/10.1093/nar/gkl950 Cho D-Y, Kim Y-A, Przytycka TM (2012) Chapter 5: network biology approach to complex diseases. PLoS Comput Biol 8(12):e1002820. https://doi.org/10.1371/journal.pcbi.1002820 Chuang H-Y, Lee E, Liu Y-T, Lee D, Ideker T (2007) Network-based classification of breast cancer metastasis. Mol Syst Biol 3:140. https://doi.org/10.1038/msb4100180 Clough E, Barrett T (2016) The gene expression omnibus database. Methods Mol Biol Clifton NJ 1418:93–110. https://doi.org/10.1007/978-1-4939-3578-9_5 Costanzo M et al (2016) A global interaction network maps a wiring diagram of cellular function. Science 353(6306):aaf1420. https://doi.org/10.1126/science.aaf1420 Danon L, Díaz-Guilera A, Duch J, Arenas A (2005) Comparing community structure identification. J Stat Mech Theory Exp 2005(09):P09008. https://doi.org/10.1088/1742-5468/2005/09/P09008 DIP:Home (n.d.). https://dip.doe-mbi.ucla.edu/dip/Main.cgi. Accessed 6 Jul 2022 Ertaylan G, Okawa S, Schwamborn JC, del Sol A (2014) Gene regulatory network - an overview. Front Cell Neurosci 8:437. https://www.sciencedirect.com/topics/neuroscience/gene- regulatory-network. Accessed 27 Jul 2022 Fabregat A et al (2017) Reactome pathway analysis: a high-performance in-memory approach. BMC Bioinformatics 18(1):142. https://doi.org/10.1186/s12859-017-1559-2 Farooqui A et al (2018) Assessment of the key regulatory genes and their interologs for turner syndrome employing network approach. Sci Rep 8(1):10091. https://doi.org/10.1038/ s41598-018-28375-0 Farooqui A, Alhazmi A, Haque S et al (2021) Network-based analysis of key regulatory genes implicated in Type 2 Diabetes Mellitus and Recurrent Miscarriages in Turner Syndrome. Sci Rep 11:10662. https://doi.org/10.1038/s41598-021-90171-0 Fornito A, Zalesky A, Bullmore E (2016) Centrality and hubs. In: Fundamentals of brain network analysis. Elsevier, pp 137–161. https://doi.org/10.1016/B978-0-12-407908-3.00005-4 Fortunato S (2010) Community detection in graphs. Phys Rep 486(3–5):75–174. https://doi. org/10.1016/j.physrep.2009.11.002 Girvan M, Newman MEJ (2002) Community structure in social and biological networks. Proc Natl Acad Sci 99(12):7821–7826. https://doi.org/10.1073/pnas.122653799 Grigorov MG (2005) Global properties of biological networks. Drug Discov Today 10(5):365–372 Guimerà R, Sales-Pardo M, Amaral LAN (2004) Modularity from fluctuations in random graphs and complex networks. Phys Rev E 70(2):025101. https://doi.org/10.1103/PhysRevE.70.025101 Guimerà R, Sales-Pardo M, Amaral LAN (2007) Classes of complex networks defined by role-to- role connectivity profiles. Nat Phys 3(1):63–69. https://doi.org/10.1038/nphys489
2 Biological Networks Analysis
47
Guzzi PH, Roy S (2020) Biological network databases. In: Biological network analysis. Academic, pp 77–93. https://www.researchgate.net/publication/341704175_Biological_network_databases. Accessed 06 Jul 2022 Haggart CR, Bartell JA, Saucerman JJ, Papin JA (2011) Whole-genome metabolic network reconstruction and constraint-based modeling. Methods Enzymol 500:411–433. https://doi. org/10.1016/B978-0-12-385118-5.00021-9 Han H et al (2018) TRRUST v2: an expanded reference database of human and mouse transcriptional regulatory interactions. Nucleic Acids Res 46(D1):D380–D386. https://doi.org/10.1093/ nar/gkx1013 Hermjakob H et al (2004) IntAct: an open source molecular interaction database. Nucleic Acids Res 32(Database issue):D452–D455. https://doi.org/10.1093/nar/gkh052 Hsu S-D et al (2011) miRTarBase: a database curates experimentally validated microRNA–target interactions. Nucleic Acids Res 39(Database issue):D163–D169. https://doi.org/10.1093/ nar/gkq1107 Hughes TR, de Boer CG (2013) Mapping yeast transcriptional networks. Genetics 195(1):9–36. https://doi.org/10.1534/genetics.113.153262 Ideker T, Ozier O, Schwikowski B, Siegel AF (2002) Discovering regulatory and signalling circuits in molecular interaction networks. Bioinformatics 18(Suppl 1):S233–S240. https://doi. org/10.1093/bioinformatics/18.suppl_1.S233 Jain S, Krishna S (2001) A model for the emergence of cooperation, interdependence, and structure in evolving networks. Proc Natl Acad Sci 98(2):543–547. https://doi.org/10.1073/ pnas.98.2.543 Jordán F, Nguyen PT, Liu W-C (2012) Studying protein-protein interaction networks: a systems view on diseases. Brief Funct Genomics 11(6):35. https://doi.org/10.1093/bfgp/els035. https://www.researchgate.net/publication/230717082_Studying_protein-protein_interaction_ networks_A_systems_view_on_diseases. Accessed 27 Jul 2022 Kandasamy K, Mohan SS, Raju R et al (2010) NetPath: a public resource of curated signal transduction pathways. Genome Biol 11:R3. https://doi.org/10.1186/gb-2010-11-1-r3. https:// genomebiology.biomedcentral.com/articles/10.1186/gb-2010-11-1-r3. Accessed 07 Jul 2022 Kanehisa M, Goto S (2000) KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28(1):27–30. https://doi.org/10.1093/nar/28.1.27 Karp PD et al (2017) The Biocyc collection of microbial genomes and metabolic pathways. Brief Bioinform 20(4):1085–1093. https://doi.org/10.1093/bib/bbx085 Kaviani S, Sohn I (2021) Complex network theory - an overview. Expert Syst Appl 180:115073. https://www.sciencedirect.com/topics/computer-science/complex-network-theory. Accessed 27 Jul 2022 Kim Y-A, Salari R, Wuchty S, Przytycka TM (2013) Module cover - a new approach to genotype- phenotype studies. Pac Symp Biocomput 2013:135–146 krzyszczyk P et al (2018) The growing role of precision and personalized medicine for cancer treatment. Technology 6(3–4):79–100. https://doi.org/10.1142/S2339547818300020 Lee D-S, Park J, Kay KA, Christakis NA, Oltvai ZN, Barabási A-L (2008) The implications of human metabolic network topology for disease comorbidity. Proc Natl Acad Sci 105(29):9880–9885. https://doi.org/10.1073/pnas.0802208105 Liu Z-P, Wu C, Miao H, Wu H (2015) RegNetwork: an integrated database of transcriptional and post-transcriptional regulatory networks in human and mouse. Database 2015:bav095. https:// doi.org/10.1093/database/bav095 MacNeil LT, Walhout AJM (2011) Gene regulatory networks and the role of robustness and stochasticity in the control of gene expression. Genome Res 21(5):645–657. https://doi. org/10.1101/gr.097378.109 Matys V et al (2003) TRANSFAC: transcriptional regulation, from patterns to profiles. Nucleic Acids Res 31(1):374–378. https://doi.org/10.1093/nar/gkg108 Medus A, Acuña G, Dorso CO (2005) Detection of community structures in networks via global optimization. Phys Stat Mech Its Appl 358(2–4):593–604. https://doi.org/10.1016/j. physa.2005.04.022
48
Najma and A. Farooqui
Milo R, Shen-Orr S, Itzkovitz S, Kashtan N, Chklovskii D, Alon U (2002) Network motifs: simple building blocks of complex networks. Science 298(5594):824–827. https://doi.org/10.1126/ science.298.5594.824 Muzio G, O’Bray L, Borgwardt K (2021) Biological network analysis with deep learning. Brief Bioinform 22(2):1515–1530. Oxford Academic. https://academic.oup.com/bib/article/22/2/1515/5964185. Accessed 27 Jul 2022 Newman MEJ (2012) Communities, modules and large-scale structure in networks. Nat Phys 8(1):25–31. https://doi.org/10.1038/nphys2162 Orchard S et al (2014) The MIntAct project—IntAct as a common curation platform for 11 molecular interaction databases. Nucleic Acids Res 42(D1):D358–D363. https://doi.org/10.1093/ nar/gkt1115 Oughtred R, Rust J, Chang C, Breitkreutz B‐J, Stark C, Willems A, Boucher L, Leung G, Kolas N, Zhang F, Dolma S, Coulombe‐Huntington J, Chatr‐aryamontri A, Dolinski K, Tyers M (2021a) BioGRID database of protein, chemical, and genetic interactions. Protein Sci 30(1):187–200. https://thebiogrid.org/. Accessed 06 Jul 2022 Oughtred R et al (2021b) The BioGRID database: a comprehensive biomedical resource of curated protein, genetic, and chemical interactions. Protein Sci Publ Protein Soc 30(1):187–200. https://doi.org/10.1002/pro.3978 Papin JA, Hunter T, Palsson BO, Subramaniam S (2005) Reconstruction of cellular signalling networks and analysis of their properties. Nat Rev Mol Cell Biol 6(2):99–111 Patra S, Mohapatra A (2020) Review of tools and algorithms for network motif discovery in biological networks. IET Syst Biol 14(4):171–189. https://doi.org/10.1049/iet-syb.2020.0004. https://pubmed.ncbi.nlm.nih.gov/32737276/. Accessed 27 Jul 2022 Pavlopoulos GA et al (2011) Using graph theory to analyze biological networks. BioData Min 4(1):10. https://doi.org/10.1186/1756-0381-4-10 Pennisi E (2005) How will big pictures emerge from a sea of biological data? Science 309(5731):94–94. https://doi.org/10.1126/science.309.5731.94 Peri S et al (2004) Human protein reference database as a discovery resource for proteomics. Nucleic Acids Res 32(Database):D497–D501. https://doi.org/10.1093/nar/gkh070 Ravasz E, Barabási A-L (2003) Hierarchical organization in complex networks. Phys Rev E Stat Nonlinear Soft Matter Phys 67(2 Pt 2):026112. https://doi.org/10.1103/PhysRevE.67.026112 Rocca-Serra P et al (2003) ArrayExpress: a public database of gene expression data at EBI. C R Biol 326(10):1075–1078. https://doi.org/10.1016/j.crvi.2003.09.026 Rubinov M, Sporns O (2010) Complex network measures of brain connectivity: uses and interpretations. NeuroImage 52(3):1059–1069. https://doi.org/10.1016/j.neuroimage.2009.10.003 Sakurai N, Ara T, Ogata Y, Sano R et al (2011) kaPPA-view4: a metabolic pathway database for representation and analysis of correlation networks of gene co-expression and metabolite co- accumulation and omics data. Nucleic Acids Res 39(Database issue):D677–D684. Oxford Academic. https://academic.oup.com/nar/article/39/suppl_1/D677/2509108?login=true. Accessed 07 Jul 2022 Schuetz P, Caflisch A (2008) Efficient modularity optimization by multistep greedy algorithm and vertex mover refinement. Phys Rev E 77(4):046112. https://doi.org/10.1103/ PhysRevE.77.046112 Sharan R et al (2005) Conserved patterns of protein interaction in multiple species. Proc Natl Acad Sci 102(6):1974–1979. https://doi.org/10.1073/pnas.0409522102 Song C, Havlin S, Makse HA (2006) Origins of fractality in the growth of complex networks. Nat Phys 2(4):275–281. https://doi.org/10.1038/nphys266 Srinivasa Rao V, Srinivas K, Sujini GN, G. N. (2014) Sunand Kumar Protein-protein interaction detection: methods and analysis. Int J Proteomics 2014:147648. https://www.ncbi.nlm.nih.gov/ pmc/articles/PMC3947875/. Accessed 27 Jul 2022 Stelzl U et al (2005) A human protein-protein interaction network: a resource for annotating the proteome. Cell 122(6):957–968. https://doi.org/10.1016/j.cell.2005.08.029 Strimbu K, Tavel JA (2010) What are biomarkers? Curr Opin HIV AIDS 5(6):463–466. https://doi. org/10.1097/COH.0b013e32833ed177
2 Biological Networks Analysis
49
Succoio M, Sacchettini R, Rossi A, Parenti G, Ruoppolo M (2022) Metabolic network - an overview. Biomol Ther 12(7):968. https://www.sciencedirect.com/topics/biochemistry-genetics- and-molecular-biology/metabolic-network. Accessed 27 Jul 2022 System-Level Organization in a Hierarchical Network Szklarczyk D et al (2019) STRING v11: protein–protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res 47(Database issue):D607–D613. https://doi.org/10.1093/nar/gky1131 Tazyeen S et al (2022) Identification of key regulators in Sarcoidosis through multidimensional systems biological approach. Sci Rep 12(1):1. https://doi.org/10.1038/s41598-022-05129-7 van den Heuvel MP, Sporns O (2011) Rich-club organization of the human connectome. J Neurosci 31(44):15775–15786. https://doi.org/10.1523/JNEUROSCI.3539-11.2011 Vieira LS, Vera-Licona P (2019) Computing signal transduction in signaling networks modeled as Boolean networks, petri nets, and hypergraphs. Biorxiv 2:272344. https://doi. org/10.1101/272344 Wee P, Wang Z (2017) Epidermal growth factor receptor cell proliferation signaling pathways. Cancer 9(5):52. https://doi.org/10.3390/cancers9050052 Wiredja D, Bebek G (2017) Identifying gene interaction networks, in statistical human genetics: methods and protocols. In: Elston RC (ed) Methods in molecular biology. Springer, New York, pp 539–556. https://doi.org/10.1007/978-1-4939-7274-6_27 Wishart DS et al (2008) DrugBank: a knowledgebase for drugs, drug actions and drug targets. Nucleic Acids Res 36(Database issue):D901–D906. https://doi.org/10.1093/nar/gkm958 Woodcock CL (2006) Chromatin architecture. Curr Opin Struct Biol 16(2):213–220. https://doi. org/10.1016/j.sbi.2006.02.005 Wu F, Huberman BA (2004) Finding communities in linear time: a physics approach. Eur Phys J B - Condens Matter 38(2):331–338. https://doi.org/10.1140/epjb/e2004-00125-x Wu M-Y, Dai D-Q, Xiao-Fei Z, Zhu Y (2013) Cancer subtype discovery and biomarker identification via a new robust network clustering algorithm. PLoS One 8:e66256. https://doi. org/10.1371/journal.pone.0066256 Wu J, Tan S-Y, Liu Z, Tan Y-J, Lu X (2017) Enhancing structural robustness of scale-free networks by information disturbance. Sci Rep 7(1):7559. https://doi.org/10.1038/s41598-017-07878-2 Yu D, Kim M, Xiao G, Hwang TH (2013) Review of biological network data and its applications. Genomics Inform 11(4):200–210. https://doi.org/10.5808/GI.2013.11.4.200 Zhang P, Itan Y (2019) Biological network approaches and applications in rare disease studies. Gene 10(10):797. https://doi.org/10.3390/genes10100797 Zhou S, Mondragon RJ (2004) The rich-club phenomenon in the internet topology. IEEE Commun Lett 8(3):180–182. https://doi.org/10.1109/LCOMM.2004.823426
3
Network Analysis Based Software Packages, Tools, and Web Servers to Accelerate Bioinformatics Research Nikhat Imam, Sadik Bay, Mohd Faizan Siddiqui, and Okan Yildirim
Abstract
Biological networks are the best way to represent complex biological systems in terms of sets of interactions between biological entities (e.g., genes, proteins, taxa, and metabolites). Now, with the availability of large-scale multi-omics data, the biological system has expanded from basic to advanced levels (like PPI connectivity and functional changes in disease stages, disease-specific interactome, genetic perturbations, and network dysfunction). In this chapter, we discuss the fundamental concept of network theory and various network types such Sadik Bay and Okan Yildrim are both currently affiliated with the Memorial Sloan Kettering Cancer Center in New York City, USA. N. Imam (*) Department of Mathematics, Institute of Computer Science and Information Technology, Magadh University, Bodh Gaya, India Centre for Interdisciplinary Research in Basic Sciences, Jamia Millia Islamia, New Delhi, India S. Bay Research Institute for Health Sciences and Technologies (SABITA), Istanbul Medipol University, İstanbul, Türkiye M. F. Siddiqui International Medical Faculty, Osh State University, Osh, Kyrgyzstan O. Yildirim Department of Chemical Biology Otto-Hahn-Strasse, Max Planck Institute of Molecular Physiology, Dortmund, Germany © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 R. Ishrat (ed.), Biological Networks in Human Health and Disease, https://doi.org/10.1007/978-981-99-4242-8_3
51
52
N. Imam et al.
as Protein–protein interaction networks, metabolic networks, genetic interaction networks, gene/transcriptional regulatory networks, and cell signaling networks. In addition, we describe network topological properties that help to understand the structure of a network which facilitates understanding the hidden mechanisms. Finally, we discuss a variety a number of online and stand-alone tools that exist for network construction, analysis, Network functional module Identification, and visualization. Overall, this chapter reaches a very broad spectrum of researchers varying from experts to beginners and they can look up a genetic universe virtually via biological networks. Keywords
Network biology · Network visualization · Tools/software
3.1 Introduction A biological network serves as a representation of complex systems, encompassing binary interactions between different biological entities, such as genes, proteins, taxa, and metabolites. The advancement of multi-omics data has expanded our understanding of biological systems, including PPI connectivity and functional changes during disease progression, disease-specific interactomes, genetic perturbations, and network dysfunction. Protein–protein interaction networks (PINs) depict the physical connections between proteins within a cellular context. In these networks, proteins are represented as nodes, and their interactions are depicted as undirected edges. PPIs play a critical role in various cellular processes and have been extensively investigated in biological research. Experimental techniques, including the yeast two-hybrid system and mass spectrometry, have been utilized to discover and identify large sets of protein interactions. In recent decades, numerous international initiatives have led to the development of databases that compile experimentally determined protein–protein interactions. Examples of these databases include the Munich Information Center for Protein Sequence (MIPS) protein interaction database (Schoof et al. 2005), the Biomolecular Interaction Network Database (BIND) (Bader 2003), the Database of Interacting Proteins (DIP) (Xenarios 2000), the Molecular Interaction database (MINT) (Chatraryamontri et al. 2007), and the protein Interaction database (IntAct) (Hermjakob 2004). These databases are classified into primary and secondary databases based on their interaction prediction method but now one new term introduced is “Metadatabase” which is a combination of different primary and secondary databases to get new and maximum protein–protein interactions in the network (Fig. 3.2). The list of protein interaction databases is given in Table 3.1.
3 Network Analysis Based Software Packages, Tools, and Web Servers to Accelerate…
53
Table 3.1 Protein–protein interaction databases Primary database Databases BioGrid
URL https://thebiogrid.org/
Experimental/ Predicted Exp.
HPRD IntAct MINT
http://www.hprd.org/ https://www.ebi.ac.uk/intact/home https://mint.bio.uniroma2.it/
Exp. Exp. Exp.
HuRI http://www.interactome-atlas.org/ Secondary database STRING https://string-db.org/
Exp.
References Oughtred et al. (2019) Goel et al. (2012) Hermjakob (2004) Chatr-aryamontri et al. (2007) Luck et al. (2020)
Exp. & pred.
Szklarczyk et al. (2019) UniHI http://www.unihi.org/ Exp. & pred. Kalathur et al. (2014) Mentha http://mentha.uniroma2.it/ Exp. Calderone et al. (2013) APID http://cicblade.dep.usal.es:8080/ Exp. Alonso-López et al. APID/init.action (2019) HIPPIE http://cbdm-01.zdv.uni-mainz. Exp. Alanis-Lobato et al. de/~mschaefer/hippie/index.php (2017) HitPredict http://www.hitpredict.org/ Exp. Patil et al. (2011) IID http://iid.ophid.utoronto.ca/ Exp. and pred. Kotlyar et al. (2016) HINT http://hint.yulab.org/ Exp. Das and Yu (2012) GPS-Prot http://gpsprot.org/ Exp. Fahey et al. (2011) • primary protein interaction databases containing literature-curated PPIs for human proteins • secondary protein interaction databases containing predicted and curated from primary databases
3.2 Types of Biological Networks There are various types of networks (Fig. 3.1). The details are given below: • Protein–protein interaction (PPI) networks. • Protein–protein interaction networks (PINs) serve as mathematical models that depict the direct interactions between proteins. These interactions, known as PPIs, play a fundamental role in nearly every cellular process, making it imperative to comprehend them in order to understand cell physiology under both normal and disease conditions. • Metabolic Networks. • Metabolic networks illustrate the relationships between enzymes and small biomolecules, referred to as metabolites, which are proteins that facilitate biochemical reactions. These networks illustrate the interactions that occur during the catalysis of these reactions.
54
N. Imam et al.
Fig. 3.1 Types of biological interactions that networks can represent
• Genetic Interaction Networks. • Genetic interaction networks provide insights into the functional relationships between pairs of genes within an organism, facilitating our understanding of the connection between genotype and phenotype. • Gene/transcriptional Regulatory Networks. • A genetic regulatory network (GRN), also known as a gene regulatory network, comprises a set of molecular regulators that interact with each other and other substances within a cell. Its purpose is to control the levels of gene expression for mRNA and proteins. • Cell Signaling Networks. • Cellular signaling networks emerge through the interaction of various cell signaling pathways and are typically identified through a combination of experimental and computational techniques.
3.3 Network Topology The properties of networks are valuable in extracting meaningful information. Network analysis aims to utilize the complexity of networks to uncover insights that would be challenging to obtain by examining individual components in isolation. Topological properties, such as the arrangement of nodes and edges within a network, are instrumental in identifying significant sub-structures. These properties can be applied to the network as a whole or to individual nodes and edges. Various topological properties and concepts are commonly used in network analysis (Fig. 3.2). • The Degree of a Network • The degree of a node indicates the number of edges connected to that specific node. This parameter is highly significant as it influences various characteristics, such as node centrality. By examining the degree distribution across all nodes in
3 Network Analysis Based Software Packages, Tools, and Web Servers to Accelerate…
55
Fig. 3.2 Topological properties of a network (figure adopted from EMBL-EBI Training)
• •
• •
• •
•
a network, we can determine whether the network exhibits a scale-free pattern, which we will explore in more detail. In visual representations, such as Fig. 3.2a, the size and color of each node correspond to its degree. In directed networks, nodes possess two types of degrees: out-degree, which represents edges leaving the node, and in-degree, which represents edges entering the node. Shortest Paths Shortest paths, which represent the minimum distance between any two nodes, provide a model for information flow within networks. This concept holds particular relevance in numerous biological networks. In Fig. 3.2b, the shortest path between nodes A and B is highlighted and spans five steps. Scale-Free Networks In scale-free networks, most nodes have connections to only a limited number of neighboring nodes. However, a small subset of high-degree nodes, known as hubs and highlighted in orange, play a crucial role in ensuring the network’s overall connectivity. This configuration is depicted in Fig. 3.2c. Transitivity Transitivity refers to the existence of closely interconnected nodes within a network, often referred to as clusters or communities. These groups of nodes exhibit stronger internal connections compared to the connections with the rest of the network, as depicted in Fig. 3.2d. Alternatively, these groups are also known as topological clusters. Centralities
56
N. Imam et al.
• Centrality measures encompass different concepts and can be assessed for both nodes and edges to determine their significance in terms of network connectivity and information flow. The degree of a node directly affects various centrality measures, such as “degree centrality,” but its influence decreases in more advanced measures like “betweenness centrality.” Figure 3.2e highlights nodes with higher betweenness centrality, indicating greater centrality, using warm colors. The size of each node in the figure corresponds to its degree.
3.4 Network Representation and Analysis Tools There are multiple tools that can be used to construct, integrate, and analyze PPI data to understand its biological meaning of the network.
3.4.1 Cytoscape It is widely regarded as a highly popular open-source software tool that facilitates the visual exploration of diverse biomedical networks, encompassing interactions involving proteins, genes, and various other types. Initially, it was designed for biological networks, but it also can be used for other purposes where show the relationship between two entities (Fig. 3.3). Cytoscape is popular among the network biologist to its working with a large variety of apps and plugins for various network analysis tailored for specific purposes, such as community search (e.g., MCODE, clusterMaker2, and JActiveModules), or for conducting Gene Set Enrichment Analysis (BiNGO, ClueGO, EnrichmentMap), are available within the Cytoscape software ecosystem.
3.4.2 GeNeCK GeNeCK is a user-friendly web server with a graphical interface, accessible at http://lce.biohpc.swmed.edu/geneck. Users can easily upload their data and submit it through the provided interface. One recommended approach for most users is to utilize ENA (Edge Neighborhood Aggregation), as it typically performs well in various scenarios without the need for manually selecting tuning parameters. ENA also provides p-values for each connection, indicating the statistical significance of the associations. Once the job is complete, the constructed network can be viewed on the GeNeCK website.
3.4.3 GeneMANIA GeneMANIA is an effective tool that utilizes a vast collection of functional association data to identify genes that are related to a given set of input genes. This comprehensive data encompasses protein and genetic interactions, pathways,
3 Network Analysis Based Software Packages, Tools, and Web Servers to Accelerate…
57
K J NODE-1 A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with interact with
NODE-2 O P Q R S T U V W X Y Z Z U A R T W R Y G B C D X A
M Z
A
O
F
Y
T
X
Q
D
L E
S
R
C W
U G N
I
P B
V
H
(Image by Dr. Veronica Espinoza (Dec 2022))
Fig. 3.3 Cytoscape is a popular tool for network analysis that represents the relationship between two entities
co-expression, co-localization, and protein domain similarity. With GeneMANIA, you can explore additional genes associated with specific pathways, complexes, or functions, such as protein kinases. The results generated by GeneMANIA depend on the genes you provide as input. For example, if your gene list represents a protein complex, GeneMANIA will suggest potential additional members of that complex. Similarly, when you input a gene list, GeneMANIA will identify connections between the genes within the selected datasets.
3.4.4 STRING STRING serves as an extensive database that encompasses a wide range of known and predicted protein–protein interactions. It covers both direct (physical) and indirect (functional) associations between proteins. The data in STRING is obtained from diverse sources, including computational predictions, knowledge transfer between organisms, and the aggregation of interactions from other primary databases. Presently, the STRING database offers valuable information on 24,584,628 proteins across 5090 organisms.
58
N. Imam et al.
3.4.5 FunCoup FunCoup utilizes a range of evidence and gold standards, including protein complexes, physical protein interactions, metabolic pathways, and signaling pathways, to predict four distinct classes of functional couplings. It integrates multiple types of evidence, such as co-expression, protein–protein interactions (PPIs), genetic interactions, PHP similarity, and co-regulation, across 11 model organisms, to construct comprehensive genome-wide networks. To facilitate the transfer of evidence between species, FunCoup relies on ortholog assignments from in paranoid.
3.5 Network Visualizer Tools 3.5.1 Arena3Dweb Arena3Dweb is the first, fully interactive and dependency-free, web application which allows the visualization of multilayered graphs in 3D space. This tool helps users to integrate multiple networks in a single view along with their intra-layer and inter-layer links. Users also can align networks and highlight the network’s topological properties, highlighting the edges and important paths. The current version supports the weighted and unweighted undirected networks and it is written in R, Shiny, and Javascript (Karatzas et al. 2021). It can be accessed by using link (https:// bib.fleming.gr:8084/app/arena3d).
3.5.2 GEPHI Gephi exhibits the ability to handle extensive networks, encompassing thousands of nodes and millions of edges, necessitating substantial computational resources. Being open source and multi-platform, it offers a diverse array of advanced network- related algorithms available as plugins, such as NET-EXPO and DyCoNet (Bastian 2009). It can be downloaded from Gephi - The Open Graph Viz Platform than can used in Windows, Mac OS X and Linux system in local machine.
3.5.3 Igraph Igraph is a versatile collection of libraries designed for graph creation, manipulation, and network analysis. Originally written in C, Igraph is also available as package for Python and R programming languages. Igraph can be installed by using R studio. Just type “install.packages(“igraph”)”.
3.5.4 Pathview The Pathview R package is a versatile toolset designed for integrating and visualizing pathway-based data. It facilitates the mapping and visualization of user data
3 Network Analysis Based Software Packages, Tools, and Web Servers to Accelerate…
59
onto pathway graphs. Users simply provide a list of gene or compound data along with the desired pathway, and Pathview takes care of automatically downloading pathway network data, parsing the data file, mapping the user data to the pathway, and generating pathway graphs with the mapped data. While Pathview can function as a standalone program, it also offers seamless integration with other pathway analysis tools, enabling large-scale and fully automated analysis pipelines (Luo and Brouwer 2013).
3.5.5 VisANT It is a specialized tool designed for visual data mining of biological networks and pathways. It offers functionalities such as integrating disease and therapy hierarchies, associations between diseases and genes, associations between therapies and drugs, and interactions between drugs and targets. The latest version of VisANT incorporates disease and drug hierarchies, disease–gene associations, therapy–drug associations, and drug–target interactions. It allows for gene and drug annotation based on disease and therapy information. Additionally, it enables the prediction of associated diseases and therapies through enrichment analysis using user-provided gene or drug sets. VisANT supports network transformation and provides a user- friendly web interface for customizing node and edge properties. It is freely available at http://visant.bu.edu (Hu et al. 2013).
3.5.6 BioNetStat This tool is accessible through the Bioconductor package in R, which provides a user-friendly graphical interface. It allows users to compare two or more correlation networks based on the probability distribution of network centrality measures. BioNetStat specializes in conducting differential network analysis, examining network features, and highlighting significant differences between disease and normal conditions using statistical significance scores. It is worth noting that this tool is not restricted to gene expression networks alone; it can also be applied to various data types such as proteomics, phenomics, metabolomics, as well as economic and social network data (Jardim 2019).
3.5.7 NetworkAnalyst The increasing use of gene expression profiling within the context of protein–protein interaction (PPI) networks requires robust and user-friendly bioinformatics tools for understanding systems-level data. This tool assists in visualizing and comparing multiple gene lists through interactive heatmaps, enrichment networks, Venn diagrams, or chord diagrams (Zhou et al. 2019). It can be accessed at https://www. networkanalyst.ca/.
60
N. Imam et al.
3.6 Network Clustering Tools 3.6.1 NeAT The Network Analysis Tools (NeAT) is a series of modular computer programs (Fig. 3.4) specifically designed for the analysis of biological networks via integration of many algorithms for the analysis of biological networks: networks comparison, clustering and pathfinding, network randomization, and network topological properties analysis (Brohée et al. 2008). NeAT user-friendly web interface to allow easy access to the tools and data processing (http://rsat.sbroscoff.fr/index_neat.html).
3.6.2 clusterMaker clusterMaker is a plugin available for Cytoscape that provides a unified interface for various clustering techniques and visualization options. It supports multiple clustering algorithms, such as hierarchical, k-medoid, AutoSOME, and k-means, which are useful for clustering expression or genetic data. Additionally, it offers partitioning algorithms like MCL, transitivity clustering, affinity propagation, MCODE, community clustering (GLAY), SCPS, and AutoSOME, enabling network partitioning based on similarity or distance values. The clustering results can be visualized as hierarchical groups of nodes or heat maps for hierarchical, k-medoid, AutoSOME, and k-means algorithms. The plugin introduces collapsible “meta nodes” that aid in interactive exploration of putative family associations within the Cytoscape network. Moreover, the results can be displayed as a separate network containing only intra-cluster edges or with inter-cluster edges included. clusterMaker is compatible with Cytoscape version 2.8.2 or newer and can be accessed through the Cytoscape plugin manager under the Analysis category.
3.6.3 GephiCrunch It utilizes a range of techniques to compare the structures of two networks, such as average clustering coefficients, average path lengths, diameters, degree distributions, clustering and eccentricity spectra, and graphlet-based heuristics. These methods enable more precise and rigorous comparisons between the networks.
3.7 Other Miscellaneous Tools 3.7.1 Network BLAST This tool enables the identification of protein complexes within protein–protein interaction networks. It has the capability to analyze either a single network or two networks from different species. In the case of analyzing two networks, Network
contingency-stats
graph-node-degree
Legend
Contingency statictics
Contingency table
Class comparison statistics
Node degrees
Statistics
PathFinder
Source and target nodes
Fig. 3.4 Network Analysis Tools (NeAT) tool map (Figure adopted from NeAT web page)
Graph drawing
display-graph
(sub-network)
Merged pathways
(sub-network)
Separate pathways
(intra / inter - cluster edges)
graph-get-clusters
Altered graph
alter-graph Labeled graph
graph-cluster-membership
Random graph
random-graph
graph-neighbours
User-selected nodes
Clustering (MCL / RNSC)
Labeled graph
String database interaction network
(intersection / union / difference)
compare-graphs
String
User-provided network
Network/graph
convet-graph
User input
method
compare-classes
Result
(membership coefficient table)
Fuzzy clusters
Node clusters
Neighbour-source node pairs
User-provided clusters
Clusters/classes
convert-classes
3 Network Analysis Based Software Packages, Tools, and Web Servers to Accelerate… 61
62
N. Imam et al.
BLAST generates a collection of complexes that exhibit evolutionary conservation across the two networks. The tool is available for download on a local Linux machine and can be installed with minimal hardware requirements (AMD XP2500+ 1 GB, Intel XEON 3 Ghz 3 GB).
3.7.2 SpectralNET A tool to analyze the networks and node interactions, such as chemical–genetic networks. It is available as a standalone. Spectral NET offers a user-friendly approach for analyzing graph-theoretic metrics in data modeling and dimensionality reduction (Forman et al. 2005).Users can conveniently access it through either a.NET application or an ASP.NET web application, available at http://chembank. broad.harvard.edu/resources/.
3.8 Conclusion A biological network is a way to represent the complex system of binary interactions or links between various biological entities. In this chapter, we have focused on Network analysis and visualization tools (Offline/Online) which make it easy for us to understand the broad maps of cellular organization. There are many network analysis visualization tools are exist, but we mainly discussed here only 11 tools that are widely used and popular among researchers. Most of the tools support a GUI (graphic user interface) where users can process tire data by a few mouse clicks, simple dialog boxes and data imports allow most functionality to be accessed. However, Several tools need some basic programing skills in R and Python to read, process, and write the data. Plugins are an important way for advanced users to customize and extend an application (Cytoscape and VisANT) based on third parties generic software to develop new functionality and integrate it directly within the tool. As we know that each tool is designed differently (different algorithm used) so the result would be different from other tools, so before using any tool, the user must be sure about what is aim of study, what type of result they want, and cross- check and validate the result with other tools. All the tools/software introduced in this chapter are freely accessible to foster collaboration among researchers and accelerate advancements in the field of systems biology.
References Alanis-Lobato G, Andrade-Navarro MA, Schaefer MH (2017) HIPPIE v2.0: enhancing meaningfulness and reliability of protein–protein interaction networks. Nucleic Acids Res 45(D1):D408–D414. https://doi.org/10.1093/nar/gkw985 Alonso-López D et al (2019) APID database: redefining protein–protein interaction experimental evidences and binary interactomes. Database 2019:baz005. https://doi.org/10.1093/ database/baz005
3 Network Analysis Based Software Packages, Tools, and Web Servers to Accelerate…
63
Bader GD (2003) BIND: the biomolecular interaction network database. Nucleic Acids Res 31(1):248–250. https://doi.org/10.1093/nar/gkg056 Mathieu Bastian, “Gephi: an open source software for exploring and manipulating networks,” Third International AAAI conference on weblogs and social media, 2009 Brohée S, Faust K, Lima-Mendez G, Vanderstocken G, van Helden J (2008) Network analysis tools: from biological networks to clusters and pathways. Nat Protoc 3(10):1616–1629. https:// doi.org/10.1038/nprot.2008.100 Calderone A, Castagnoli L, Cesareni G (2013) Mentha: a resource for browsing integrated protein- interaction networks. Nat Methods 10(8):690–691. https://doi.org/10.1038/nmeth.2561 Chatr-aryamontri A et al (2007) MINT: the molecular INTeraction database. Nucleic Acids Res 35, no. Database:D572–D574. https://doi.org/10.1093/nar/gkl950 Das J, Yu H (2012) HINT: high-quality protein interactomes and their applications in understanding human disease. BMC Syst Biol 6(1):92. https://doi.org/10.1186/1752-0509-6-92 Fahey ME et al (2011) GPS-Prot: a web-based visualization platform for integrating host-pathogen interaction data. BMC Bioinformatics 12(1):298. https://doi.org/10.1186/1471-2105-12-298 Forman JJ, Clemons PA, Schreiber SL, Haggarty SJ (2005) SpectralNET–an application for spectral graph analysis and visualization. BMC Bioinformatics 6(1):260. https://doi.org/10.118 6/1471-2105-6-260 Goel R, Harsha HC, Pandey A, Prasad TSK (2012) Human protein reference database and human Proteinpedia as resources for phosphoproteome analysis. Mol BioSyst 8(2):453–463. https:// doi.org/10.1039/C1MB05340J Hermjakob H (2004) IntAct: an open source molecular interaction database. Nucleic Acids Res 32(90001):452D–4455D. https://doi.org/10.1093/nar/gkh052 Hu Z et al (2013) VisANT 4.0: integrative network platform to connect genes, drugs, diseases and therapies. Nucleic Acids Res 41, no. Web Server issue:W225–W231. https://doi.org/10.1093/ nar/gkt401 Jardim VC, Santos S d S, Fujita A, Buckeridge MS (2019) BioNetStat: a tool for biological networks differential analysis. Front Genet 10:594. https://doi.org/10.3389/fgene.2019.00594 Kalathur RKR et al (2014) UniHI 7: an enhanced database for retrieval and interactive analysis of human molecular interaction networks. Nucleic Acids Res 42(D1):D408–D414. https://doi. org/10.1093/nar/gkt1100 Karatzas E, Baltoumas FA, Panayiotou NA, Schneider R, Pavlopoulos GA (2021) Arena3Dweb: interactive 3D visualization of multilayered networks. Nucleic Acids Res 49(W1):W36–W45. https://doi.org/10.1093/nar/gkab278 Kotlyar M, Pastrello C, Sheahan N, Jurisica I (2016) Integrated interactions database: tissue-specific view of the human and model organism interactomes. Nucleic Acids Res 44(D1):D536–D541. https://doi.org/10.1093/nar/gkv1115 Luck K et al (2020) A reference map of the human binary protein interactome. Nature 580(7803):402–408. https://doi.org/10.1038/s41586-020-2188-x Luo W, Brouwer C (2013) Pathview: an R/Bioconductor package for pathway-based data integration and visualization. Bioinformatics 29(14):1830–1831. https://doi.org/10.1093/ bioinformatics/btt285 Oughtred R et al (2019) The BioGRID interaction database: 2019 update. Nucleic Acids Res 47(D1):D529–D541. https://doi.org/10.1093/nar/gky1079 Patil A, Nakai K, Nakamura H (2011) HitPredict: a database of quality assessed protein–protein interactions in nine species. Nucleic Acids Res 39, no. suppl_1:D744–D749. https://doi. org/10.1093/nar/gkq897 Schoof H et al (2005) Munich information Center for Protein Sequences Plant Genome Resources. A framework for integrative and comparative analyses. Plant Physiol 138(3):1301–1309. https://doi.org/10.1104/pp.104.059188 Szklarczyk D et al (2019) STRING v11: protein–protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res 47(D1):D607–D613. https://doi.org/10.1093/nar/gky1131
64
N. Imam et al.
Xenarios I (2000) DIP: the database of interacting proteins. Nucleic Acids Res 28(1):289–291. https://doi.org/10.1093/nar/28.1.289 Zhou G, Soufan O, Ewald J, Hancock REW, Basu N, Xia J (2019) NetworkAnalyst 3.0: a visual analytics platform for comprehensive gene expression profiling and meta-analysis. Nucleic Acids Res 47(W1):W234–W241. https://doi.org/10.1093/nar/gkz240
4
Networks Analytics of Heterogeneous Big Data Rafat Ali and Nida Jamil Khan
Abstract
Day by day exponential growth in experimentation is generating staggering amounts of data in the different fields of biological research. Generating this high volume of data brought the term “Big Data.” The term “Big Data” has several dimensions which contain a huge amount of unstructured and structured data sets. Analysis, interpretation, and complete extraction of meaningful information from the raw, structured, and unstructured datasets lead to big scientific inventions, open the routes of progress in industry and economic development. Due to bulky data set makes it is very complex and complicated in the sense of co- relationship and connections among them, managing the hierarchy level, and many data linkages. It is extremely difficult to easily manage data quality with proper security and privacy with such a large number of multivariate raw data. Since “Big Data” encompasses both organized and unstructured data sets, therefore storing, transferring, processing, and searching the raw data is extremely challenging, and it cannot be managed using traditional database systems and software tools. The heterogeneous “Big data” set contains petabytes or exabytes of raw data, with billions to trillions of archived records. To gain hidden information from the big, archived records complex network biology approach has a significant role. Biological complex networks like gene–gene and protein–protein interaction networks have been appreciated for finding genes and pathways associated with diseases. These complex networks could provide significant insights into the mechanisms of complex diseases like cancer.
R. Ali · N. J. Khan (*) Department of Biosciences, Jamia Millia Islamia, New Delhi, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 R. Ishrat (ed.), Biological Networks in Human Health and Disease, https://doi.org/10.1007/978-981-99-4242-8_4
65
66
R. Ali and N. J. Khan
Keywords
Big data · Network analytics · Genetic profile · Next generation sequencing · Data quality · Genomic databases
4.1 Big Data Analytics for Network Biology In the various domains of biological study, exponential increase in experimentation has produced huge amounts of data. The phrase “Big Data” was coined to describe the emerging field of modern biology due to the huge amount of data generated. The phrase “Big Data” refers to a number of dimensions that include a sizable number of unstructured and organized data sets. Big scientific discoveries are made possible by the analysis, interpretation, and thorough extraction of useful data from large raw, structured, and unstructured databases. This also paves the way for advancements in business and economic growth. Because of the data set’s weight, managing the hierarchy level and numerous data links, as well as its co-relationships and linkages, makes it extremely difficult. Easily managing such a massive volume of multivariate raw data is challenging. Because “Big Data” includes both structured and unstructured data sets, managing the raw data is exceedingly difficult. It cannot be done with the help of conventional database management systems and software tools. The heterogeneous “Big data” collection might include trillions of old records together petabytes or exabytes in size (Altaf-Ul-Amin et al. 2014). The era of Big Data is closely related to the era of Omics, and omics technologies and assays, in particular, omics technologies and assays are key to producing enormous amounts of data. So, omics is a significant stakeholder in big data. Future generations will use emerging technologies like imaging systems, spectroscopy-based flow cytometry, different sequencing techniques, and others to produce enormous amounts of data (Nielsen et al. 2010). Big challenges of big data in biology to store, analyze, and interpret have been brought in by the greater volume of nucleic acid sequencing data and output of advanced high-throughput techniques in system biology, whether it is cell biology or molecular biology, producing piles of omics data in different omics-based fields like-genomes, transcriptomes, proteomes, metabolomes, interactomes, and so on (Pal et al. 2020). Additional sources that are expanding exponentially are also producing a significant amount of raw data. Analysis of a big amount of data is really a concern of the new scientific era. Big data analytics is a fast-growing technology that has been used in various areas, including network biology. The complex network biology approach plays a vital role in extracting hidden information from large stored records. To uncover disease-associated genes and pathways complex networks such as gene–gene or protein–protein interaction networks have gained importance. It is possible that these intricate networks may reveal important information about how complicated diseases like cancer function. Since every gene, protein, piece of DNA, and pathway in the human body is somehow related to one another, a complex network approach offers a simulation method for learning actual information about the gene, protein, etc. plays a vital role in extracting hidden information from the big stored biological data (Fig. 4.1).
4 Networks Analytics of Heterogeneous Big Data
67
4.2 Genetic Profiling Data The word “genetic profile” is typically used to refer to a composition of genetic traits that are associated with a human individual, such as genetic signals or information. It is crucial to emphasize that profiling is the method by which such a collection of traits is linked to the target qualities considered when making decisions, such as any disease occurrence (European Commission 2018). In healthcare and scientific research, genetic profiling is used to link particular genetic traits to increased or decreased risk of detecting and curing specific diseases. Rarely diseases are monogenic or caused by genetic variations at a single locus that have a significant impact on the disease’s course. An illustration of such a monogenic disease is cystic fibrosis. Genetic diseases are often polygenic, which means they are probably caused by a combination of various genes, lifestyle choices, and environmental factors. Even though several DNA-related markers are important for identifying individuals and particular characteristics, in general, only a very small portion of DNA is important for medical treatment and study, such as genes, SNPs, Short Tandem Repeats (STR), and whole genome sequences. In contrast to, RNA variations, these markers are directly associated with the DNA without any translation step in between, allowing one to concentrate on stable molecular properties. Because STRs are typically employed to create an individual’s genetic fingerprint, they are extremely important (Sariyar et al. 2017). Profiling may be done in various situations, including forensic science, marketing, healthcare settings, and information science. Big data expands how profiling may be used to find new patterns and make automated decisions (Hildebrandt 2008). Healthcare profiling allows for more tailored decisions than those based on average traits, which is closely connected to personalized therapy. Its potential applications include prognostication, resistance detection, disease monitoring, risk assessment, recurrence
Fig. 4.1 Figure illustrating big data handling and analytical approach
68
R. Ali and N. J. Khan
detection, and early detection. Biomarker-guided diagnostic decisions and therapy selection are significant (Wjst 2010).
4.3 Data Quality By 2025, it is estimated that between 100 million and 2 billion human genomes will have been sequenced, making genomics the world’s biggest “big data” challenge (Stephens et al. 2015). Increasing amounts of genomic data of various types have been generated by high-throughput technologies and, more recently, Next Generation Sequencing (Schuster 2008), making significant advancements in the understanding of human genome mechanisms and their application to previously unheard-of personalized medicine outcomes. Data and metadata integration is regarded as an activity of unquestionable importance prior to data analysis and biological knowledge discovery, with pressing demands for improved data extraction, matching, normalization, and enrichment methodologies to enable building multiple perspectives over the genome; these can lead to the identification of meaningful relationships, which are otherwise not perceptible when using incompatible data representations (Samarajiwa et al. 2018). The way that bioinformatics and genomics, in past years operated by utilizing the extensive fieldwork done by its practitioners in terms of data collection, management, and analysis. Best practices are gathered from many laboratories and projects, discussed on discussion boards (like https://www.biostars. org, http://seqanswers.com, and https://www.researchgate.net), and compiled in documentation or wiki guides in tool and software source repositories. The quality of the experimental data generated by these procedures, for which standard protocols are frequently made up of several scripts, are accessible, and become a primary interest to bioinformaticians. In contrast, less focus has been placed on the high- quality actions that may be taken when combining several experimental data in systematic methods. Handling data standards in both schemata and values, as well as applying integration methods that enhance data quality, become more crucial with the development of a culture of data FAIRness (Wilkinson et al. 2016) and of open and sharable research, which is supported by efforts like FAIRsharing (Sansone et al. 2019) focusing on accuracy, consistency, currency, and reliability (Hedeler and Missier 2008). One out of three business leaders do not trust the data they use to make decisions, according to IBM’s 2012 report on data quality (https://opensistemas.com/en/the-four-vs-of-big-data); this ratio is unacceptable in industries like healthcare and precision medicine, which are heavily reliant on genomic databases and decision-making techniques. “Quality” in genomics has typically been used to refer to “quality control” steps on sequences, typically a pre-processing activity aimed at removing adapter sequences, low-quality reads, uncalled bases, and contaminants. This usage dates back to DNA microarrays (Ji and Davis 2006) and Next Generation Sequencing (Schuster 2008). Instead, we use the term “Data Quality” (DQ) in this study to refer to the larger definition given by Wang and Strong (Wang and Strong 1996), which is typically encapsulated by the phrase “fitness for use,” i.e., the capacity of datasets
4 Networks Analytics of Heterogeneous Big Data
69
to satisfy the needs of their users. DQ is assessed using many quality factors (i.e., single aspects or components of a data quality concept (Stvilia et al. 2007). A summary of cutting-edge methods for resolving problems with data quality in general databases is given in (Fan 2015), under the heading “data cleaning.” Data of low quality in genetic databases have a huge economic and medical impact on their users/customers. For instance, incorrect target selection for medical investigations or pharmaceutical research may result from inaccuracies in genomic data. Pharmaceutical corporations invest billions of dollars in research just to release a few novel medications (Hensley 2002). Only a small number of the hundreds of potential leads generated from experimental genetic data make it to clinical trials, and only one medicine is commercially viable. It is of considerable importance to base these far-reaching judgments on good-quality data (Shankaranarayanan et al. 2000).
4.4 Major Public Databases Genome mapping and sequencing technologies are producing a huge amount of data; as a result, dependable and effective storage methods are needed. The first organization of genome sequencing and mapping data involved manual data collection and deposit in a central location, such as a table with columns and rows. This uses resources in a laborious, time consuming, and ineffective manner. A drawback was rapidly identified as the absence of synergy between big independent collections of different sorts of data. Data are combined through the science of informatics developed by information technology experiments in the lab that enables the gathering, analysis, and distribution of beneficial combinations of pertinent data. In order to minimize mistakes, eliminate redundancy across comparable data sources, and make well-informed judgments regarding outcomes, informatics offers a semi- automated method for retrieving, filtering, and making comparisons and contrasts of data in an electronic format. Many databases are created using tables or relational databases as their foundation. Alternatives include object-oriented techniques that allow for flexible data categorization and analysis based on this classification, as well as data storage and retrieval (Carroll et al. 2001). Biology is undergoing a revolution as a result of genome-wide analyses of gene function and expression as well as a genomic structure made possible by genomic sequence data. The application of human genetic data is anticipated to have profound effects on pathology and the creation of individualized treatments. Online genomic databases provide free access to genome reference sequences for thousands of species. Genome browser’s user-friendly software that produces interactive, graphical representations of key chromosomal areas with comprehensive annotations, including genes, epigenetic information, and sequence variations, may be used to immediately download or search sequence data. In addition to detailing several methods for searching them, such as employing IDs for genes and chemicals, karyotype bands, chromosomal locations, sequences, and motifs, this chapter enlisted Table 4.1 major genomic databases and genome browsers. The human
70
R. Ali and N. J. Khan
genome is highlighted, along with methods for visualizing and retrieving data related to genome plasticities, such as sequence and structural variations (Hutchins 2020). As the principal public repository for genomic sequence data, the National Center for Biotechnology Information (NCBI) gathers and preserves vast quantities of diverse data. The NCBI website’s text-based search and retrieval system, which enables quick and simple access to a variety of biological databases, integrates data on genomes, genes, gene expressions, gene variation, gene families, proteins, and protein domains with analytical, search, and retrieval capabilities. The speed of discovery is accelerated by the use of comparative genomic analysis methods to gain a deeper knowledge of evolutionary processes. Our knowledge of the biology of living things has undergone a fundamental transformation as a result of the increase in genome sequencing that has been sparked by recent technological advancements. The information management system and the visualization tools will now face additional difficulties as a result of the enormous growth in DNA sequence data. To organize this genomic sequence flood and enhance the usefulness of the related data, new approaches need to be developed (Tatusova 2016).
4.5 Challenges of Handling Genomics’ Data We reported here a number of difficulties that data scientists use open datasets to address biological and clinical issues confront. Despite the significant effort put into creating such public datasets, they are spread across several sources, varied in terms of their formats, and frequently fulfill extremely varying quality criteria (Ceri and Pinoli 2020). The development of high-throughput technologies at much lower prices has led to a massive data explosion in genomics in recent years. The age of millions of accessible genomes is about to begin. Notably, each genome can include billions of nucleotides that are encoded in terabytes of plain text (GBs). It is obvious that these genetic data provide unaware of data difficulties (Wong 2019). There has been an explosion of genetic data created as a result of recent advancements in biotechnology, particularly next generation sequencing in genomics. Both in terms of volume and variety, the findings are substantial. Much more information is contained in big data, which also presents data analysis with novelty. The high- throughput data produced by NGS and similar technologies is called big data. Volume, Velocity, Variety, Veracity, and Value, sometimes known as the 5Vs, are characteristics of Big Data that make it distinctive in terms of difficulties and potential. Big Data in genomics is being produced at a pace and scale never before seen because of new technologies like NGS. Data analysis is challenging because of Big Data. Regarding the study of Big Data in genomics, we talked about the difficulties with data integration, data administration, computer infrastructure, dimension reduction, data smoothing, and data security. The analysis of single-cell sequencing data, de-novelization of sequencing reads, and the study of rare genetic variations are a few more issues that are more data specific. These fields are the subject of active research and will deliver crucial data and tools for comprehending genomics
Resource NCBI resources RefSeq homepage Genomes homepage Genome data download (FTP) Human genome page Genome Data Viewer (GDV) GDV tutorial Viral genomes Ensembl Homepage Genome data download (FTP) Human genome page Ensembl genomes Ensembl bacteria Ensembl fungi Ensembl metazoa Ensembl plants Ensembl protists Ensembl tutorials UCSC Genome Browser Homepage Genome data download (FTP) Genome Browser User Guide Stand-alone genome browsers Integrated Genome Browser (IGB) Integrative Genomics Viewer (IGV) NCBI Genome Workbench
https://bioviz.org/ http://software.broadinstitute.org/software/igv/ https://www.ncbi.nlm.nih.gov/tools/gbench/
https://genome.ucsc.edu/ http://ftp//hgdownload.soe.ucsc.edu/goldenPath/currentGenomes/ https://genome.ucsc.edu/goldenPath/help/hgTracksHelp.html
https://www.ensembl.org/ http://ftp//ftp.ensembl.org/pub/ https://www.ensembl.org/Homo_sapiens/Info/Index http://ensemblgenomes.org/ https://bacteria.ensembl.org/ https://fungi.ensembl.org/ https://metazoa.ensembl.org/ https://plants.ensembl.org/ https://protists.ensembl.org/ https://www.ensembl.org/info/website/tutorials/index.html
https://www.ncbi.nlm.nih.gov/refseq/ https://www.ncbi.nlm.nih.gov/genome http://ftp//ftp.ncbi.nlm.nih.gov/genomes/ https://www.ncbi.nlm.nih.gov/projects/genome/guide/human/index.shtml https://www.ncbi.nlm.nih.gov/genome/gdv/ https://www.ncbi.nlm.nih.gov/genome/gdv/browser/help/ https://www.ncbi.nlm.nih.gov/genome/viruses/
Website
Table 4.1 Genomic databases and genome browsers
Freese et al. (2016) Thorvaldsdottir et al. (2013)
Haeussler et al. (2019)
Kersey et al. (2018)
Brister et al. (2015) Cunningham et al. (2019)
References Sayers et al. (2019) O’Leary et al. (2016)
4 Networks Analytics of Heterogeneous Big Data 71
72
R. Ali and N. J. Khan
and life in general (Xu 2020). High dimensionality, which relates to both the sample size and the number of variables and their structures, distinguishes big data in genomics. The sheer amount of data presents problems for computing and data storage. For only the raw data of each sample, the volume of data might be terabytes. It is a good idea to maintain the raw data for the various forms of genomic data, frequently in the form of picture files, so that when more advanced base calling algorithms become available, they may be used to increase accuracy (Barrett et al. 2013). The computation in genomics is impractical with current computer infrastructure due to the massive amounts of Big Data. For research with huge sample sizes, it can take months to complete the alignment and annotation of NGS reads utilizing desktop PCs. Utilizing high-performance computing resources like computer clusters is one way to solve this issue. The goal is to divide the large computing task into smaller tasks and distribute them across the cluster’s compute nodes. The result is highly parallel computing, which enables quick completion of large tasks (Almasi and Gottlieb 1989). In the field of processing large data, there are several problems that need to be solved. These problems need to be solved holistically, utilizing the knowledge of several computer science disciplines. Data Cleansing/Acquisition/ Capture, Data Storage/Sharing/Transfer, Data Analysis, as well as certain ethical considerations that come from the exposure and processing of large data, are some big data management challenges that scientists should address (Jagadish et al. 2014).
4.6 Security of Genomics’ Data A wealth of genetic data has recently been generated due to the lower cost of DNA sequencing, which is used to promote scientific research, enhance clinical practices, and enhance healthcare delivery. Genome-wide association studies (GWASs), diagnostic testing, personalized medicine, and drug discovery are all being revolutionized by these developments. The human genome is complex in nature and uniquely identifies an individual. Therefore this presents security and privacy problems. In this chapter, we discuss the issue of genomic privacy and evaluate pertinent privacy assaults that have been utilized to invade someone’s privacy. These attacks may be categorized as identity tracking, attribute disclosure, and completeness attacks (Abukari and Chen 2020). The development of genome sequencing methods promotes the accessibility of genetic data and its gathering for processing, sharing, and storage. Despite these advantages, there are still significant worries regarding how genetic data is stored, shared, transported, and processed. DNA donors occasionally inquire about the storage of their genetic information. Who can access it? What are safety precautions in place to safeguard my privacy? These worries result from the fact that (Australian Genomics 2016) Applications of genetic data now and in the future might lead to ethical and privacy issues. The danger of misuse by prospective criminals is increased by the processing and storing of this data since the human genome contains sensitive personal data.
4 Networks Analytics of Heterogeneous Big Data
73
4.7 Conclusion We are currently in the era of “big data,” in which big data technology is being rapidly applied to biomedical and healthcare fields. We might face big problems if we do not have big solutions for big amounts of data. With sophisticated platforms and tools like gene sequencing mapping tools now in use to assist in analyzing biological data, big data application in bioinformatics is still in a relatively early stage of development. To analyze big data network biology approach is a big solution. However, before analyzing the big amount of data, researchers need to improve several things. The absence of consistency in laboratory practices and values makes it difficult to integrate data. For instance, imaging data that originates from many laboratories using various techniques can experience technological batch effects. When there is a batch effect, efforts are made to normalize the data; while this may be simpler for image data, normalizing laboratory test data is inherently more challenging. Big data integration and usage in all domains continue to be hampered by security and privacy issues; as a result, secure platforms with improved communication standards and protocols are urgently required. Big Data involves complex systems, revenue, and difficulties. To increase the effective evaluation online as well as the display, analysis, and storage of Big Data, more research is required. Big Data’s main security issues are privacy, integrity, availability, and confidentiality of outsourced data.
References Abukari MY, Chen Y-PP (2020) Ensuring privacy and security of genomic data and functionalities. Brief Bioinform 21(2):511–526 Almasi GS, Gottlieb A (1989) Highly parallel computing. Benjamin-Cummings Publishing Co., Inc., Redwood City Altaf-Ul-Amin M et al (2014) Systems biology in big data and networks. Biomed Res Int 2014:428570 Australian Genomics. A National approach to data federation and analysis. 2016 Barrett T, Wilhite SE, Ledoux P, Evangelista C, Kim IF, Tomashevsky M, Marshall KA et al (2013) NCBI GEO: archive for functional genomics data sets—update. Nucleic Acids Res 41:D991–D995 Brister JR, Ako-Adjei D, Bao Y, Blinkova O (2015) NCBI viral genomes resource. Nucleic Acids Res 43(Database issue):D571–D577 Carroll ML, Nguyen SV, Batzer MA (2001) Genome databases. In: e LS Ceri S, Pinoli P (2020) Data science for genomic data management: challenges, resources, experiences. SN Computer Science 1(1):1–7 Cunningham F, Achuthan P, Akanni W et al (2019) Ensembl 2019. Nucleic Acids Res 47(D1):D745–D751 European Commission (2018). Guidelines on automated individual decision-making and profiling for the purposes of regulation 2016/679 (wp251rev.01) Fan W (2015) Data quality: from theory to practice. ACM SIGMOD Rec 44(3):7–18 Freese NH, Norris DC, Loraine AE (2016) Integrated genome browser: visual analytics platform for genomics. Bioinformatics 32(14):2089–2095 Haeussler M, Zweig AS, Tyner C et al (2019) The UCSC genome browser database: 2019 update. Nucleic Acids Res 47(D1):D853–D858
74
R. Ali and N. J. Khan
Hedeler C, Missier P (2008) Information quality management challenges for high-throughput data. In: Biological database model, p 81 Hensley S (2002) Death of Pfizer’s ‘youth pill’ illustrates drug makers woes. The Wall Street Journal online Hildebrandt M (2008) Defining profiling: a new type of knowledge? In: Hildebrandt M, Gutwirth S (eds) Profiling the European citizen: cross-disciplinary perspectives. Springer Netherlands, Dordrecht, pp 17–45. https://doi.org/10.1007/978-1-4020-6914-7_2 Hutchins JR (2020) Genomic databases. In: Genome plasticity in health and disease. Academic Press, pp 47–62 Jagadish HV, Gehrke J, Labrinidis A, Papakonstantinou Y, Patel JM, Ramakrishnan R, Shahabi C (2014) Big data and its technical challenges. Commun ACM 57(7):86–94 Ji H, Davis RW (2006) Data quality in genomics and microarrays. Nat Biotechnol 24(9):1112–1113 Kersey PJ, Allen JE, Allot A et al (2018) Ensembl genomes 2018: an integrated omics infrastructure for nonvertebrate species. Nucleic Acids Res 46(D1):D802–D808 Nielsen CB et al (2010) Visualizing genomes: techniques and challenges. Nat Methods 7(3):S5–S15 O’Leary NA, Wright MW, Brister JR et al (2016) Reference sequence (RefSeq) database at NCBI: current status, taxonomic expansion, and functional annotation. Nucleic Acids Res 44(D1):D733–D745 Pal S et al (2020) Big data in biology: the hope and present-day challenges in it. Gene Reports 21(4):100869 Samarajiwa SA, Olan I, Bihary D (2018) Challenges and cases of genomic data integration across technologies and biological scales. In: Advanced data analytics in health. Springer, pp 201–216 Sansone S-A, McQuilton P, Rocca-Serra P, Gonzalez-Beltran A, Izzo M, Lister AL, Thurston M (2019) Fairsharing as a community approach to standards, repositories and policies. Nat Biotechnol 37(4):358–367 Sariyar M, Suhr S, Schlünder I (2017) How sensitive is genetic data? Biopreserv Biobank 15:494–501. https://doi.org/10.1089/bio.2017.0033 Sayers EW, Agarwala R, Bolton EE et al (2019) Database resources of the national center for biotechnology information. Nucleic Acids Res 47(D1):D23–D28 Schuster SC (2008) Next-generation sequencing transforms today’s biology. Nat Methods 5(1):16–18 G. Shankaranarayanan, R.Y. Wang, M. Ziad, IP-MAP: representing the manufacture of an information product. In proceedings of the International conference on information quality (IQ), Cambridge, 2000, 1–16 Stephens ZD, Lee SY, Faghri F, Campbell RH, Zhai C, Efron MJ, Iyer R, Schatz MC, Sinha S, Robinson GE (2015) Big data: astronomical or genomical? PLoS Biol 13(7):e1002195 Stvilia B, Gasser L, Twidale MB, Smith LC (2007) A framework for information quality assessment. J Am Soc Inf Sci Technol 58(12):1720–1733 Tatusova T (2016) Update on genomic databases and resources at the national center for biotechnology information. In: Data mining techniques for the life sciences. Humana Press, New York, pp 3–30 Thorvaldsdottir H, Robinson JT, Mesirov JP (2013) Integrative genomics viewer (IGV): high- performance genomics data visualization and exploration. Brief Bioinform 14(2):178–192 Wang RY, Strong DM (1996) Beyond accuracy: what data quality means to data con-sumers. J Manag Inf Syst 12(4):5–33 Wilkinson MD, Dumontier M, Aalbersberg IJ, Appleton G, Axton M, Baak A, Blomberg N, Boiten J-W, da Silva Santos LB, Bourne PE et al (2016) The fair guiding principles for scientific data management and stewardship. Sci Data 3:160018 Wjst M (2010) Caught you: threats to confidentiality due to the public release of large-scale genetic data sets. BMC Med Ethics 11:21. https://doi.org/10.1186/1472-6939-11-21 Wong KC (2019) Big data challenges in genome informatics. Biophys Rev 11(1):51–54 Xu H (2020) Big data challenges in genomics. In: Handbook of statistics, vol 43. Elsevier, pp 337–348
5
Network Medicine: Methods and Applications Aftab Alam, Okan Yildirim, Faizan Siddiqui, Nikhat Imam, and Sadik Bay
Abstract
Network medicine (NM) is a developing field within network science that focuses on molecular and genetic interrelationships, disease network biomarkers, and the discovery of therapeutic targets, and it is a rapidly growing arena for medical science and research, which comes with the possibilities to reform the system of disease diagnosis and its treatment. The NM uses topological and dynamic properties of the biological networks (protein–protein interactions and metabolic pathways) to distinguish the disease patterns (characterizing the behavior of disease genes) and associated drugs. Biomedical data provide a base to develop a significant model and get potential results at the network level. In Aftab Alam has become affiliated with the United Arab Emirates University, UAE. Okan Yildirim and Sadik Bay are both currently affiliated with the Memorial Sloan Kettering Cancer Center in New York City, USA. A. Alam Centre for Interdisciplinary Research in Basic Sciences, Jamia Millia Islamia, New Delhi, India O. Yildirim Department of Chemical Biology Otto-Hahn-Strasse, Max Planck Institute of Molecular Physiology, Dortmund, Germany F. Siddiqui International Medical Faculty, Osh State University, Osh, Kyrgyzstan N. Imam Department of Mathematics, Institute of Computer Science and Information Technology, Magadh University, Bodh Gaya, India S. Bay (*) Research Institute for Health Sciences and Technologies (SABITA), Istanbul Medipol University, Istanbul, Turkey © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 R. Ishrat (ed.), Biological Networks in Human Health and Disease, https://doi.org/10.1007/978-981-99-4242-8_5
75
76
A. Alam et al.
this chapter, we have discussed the integrative use of emerging tools and the databases of NM, which provide a basic platform to systematically investigate the molecular complexity of diseases, dominant disease genes (modules), drug targets, disease-enriched pathways (altered pathways), and molecular interactions between apparently distinct phenotypes. The field of network medicine and its implications for diagnostics, prognosis, and therapeutics with unprecedented breadth and precision have great potential in the future. Keywords
Network medicine · System pharmacology · Precision medicine · Disease–gene relationship · Drug repurposing
5.1 Introduction “Interaction” is a single word that makes communication among the physical body (human beings), and even cellular components activate its functional switches by “interactions” with another component of the cell; overall, these interactions represent the human interactome. Interactome assessment is exclusively a tool-based approach to the theoretical paradigm and methodological tools utilized, describe, investigate, and comprehend structural and relational features of human health and disorders. Network-based research is becoming a crucial technique for identifying disease susceptibility genes and their associations with various diseases (Alam et al. 2022). Additionally, this research has enhanced our comprehension of drug targets and their results and proposed new drug targets, treatments, and therapeutic management strategies for serious disorders (Fig. 5.1). Network-based approaches to human disease offer a variety of biological and therapeutic uses. Indeed, a better comprehension of the consequences of (1) cellular interconnections failure or (2) rewiring in cellular interconnections on disease progression could lead to the identification and classification of disease-associated genes and pathways which could provide better targets for drug development. Advancement in these fields could also restructure the clinical application and practice, from the development of improved and more precise biomarkers that monitor the functional integrity of the network that is perturbed by the diseases as well as improvements in disease classification and pave the way to personalized therapies and treatment. In a recent study, Gysi et al. employed network medicine and drug repurposing methods to distinguish the repurposable drugs for COVID-19 (Morselli Gysi et al. 2021). Similarly, a multi-target herb called Caesalpinia pulcherima (CP) is used therapeutically to treat breast cancer (Sakle et al. 2020). Furthermore, Azuaje et al. contributed to our understanding of the cardiovascular effects of non-cardiovascular medications by integrating multiple sources of drug and target interaction data. They constructed the myocardial infarction (MI) drug-target interactome network, offering systemic insights into the topic (Azuaje et al. 2011). In a related study, Kim et al. have proposed that examining the network-based drug–disease intimacy can offer a novel perspective on the therapeutic effects of drugs in the context of
5 Network Medicine: Methods and Applications
77
Fig. 5.1 Overview of network medicine approach
Systemic Sclerosis (SSc) disease. This approach may provide insights into drug combinations or drug repositioning strategies (Kim et al. 2020). In this chapter, we focused on networks and their role in disease, System Pharmacology, Pharmacogenomics in Precision Medicine, Drug–Target Interaction, Drug–Drug Interaction, Drug Repurposing Opportunities, Drug Side Effects, and Integrating Omics data with Networks: Challenges and Ways. We hope this new chapter provides the same platform for scholars from biological science backgrounds to work on interdisciplinary research areas that can be useful for describing the causes of disease and pinpointing potential treatment targets, which will improve preventative healthcare and have a knock-on effect on personalized therapy.
5.2 Basic Principles and Key Components of Network Medicine 5.2.1 Systems Pharmacology The early understanding of the molecular mechanisms behind pharmacological action was provided by classical investigations, such as the development of the receptor hypothesis that distinguished between competitive and noncompetitive inhibition. The prevalence of medications that target membrane receptors (mostly GPCRs) explains the influence and applicability of receptor theory in contemporary
78
A. Alam et al.
pharmacology. The majority of the remaining drugs are enzyme inhibitors, which are often analogs based on substrates, analogs based on transition states, or allosteric inhibitors that are designed to bind reversibly or irreversibly based on the substrate and substrate-binding pocket configuration. The genomic, proteomic, network, and other high-throughput investigations have produced a wealth of “systems-level” knowledge over the last 10 years. The number of well-characterized, druggable targets is continuously rising through the use of high-throughput technology, structural and biochemical studies, and human genome research. Targeted therapies and biological medicines have thus emerged as a result of the expansion of the pharmacological pipeline and concentration on complicated, multigenic disorders. In recent studies, regulatory network analysis and structural analysis were used to anticipate the therapeutic benefits of medications for complicated disorders as well as potential off-target consequences (Boran and Iyengar 2010; Black and Leff 1983; Maehle et al. 2002; Colquhoun 2006).
5.2.2 Pharmacogenomics in Precision Medicine One of the fundamental components of personalized treatment is pharmacogenomics (PGx). In personalized medicine, also known as precision medicine, patients are given prescriptions for drugs that are right for them based on their genetic, environmental, and lifestyle characteristics. Two key functions of pharmacogenomics in precision medicine. It first directs pharmaceutical firms in drug development and discovery. Second, it helps doctors choose the best medication for patients based on their genetic makeup, avoid adverse drug reactions, and maximize drug efficacy by providing the appropriate amount. Based on the understanding of pharmacogenomics, personalized/precision medicine has significant potential benefits. Precision medicine is the future of healthcare, and it will eventually become the standard of care.
5.2.3 Biological Networks and Important Databases A method of expressing systems as complicated sets of binary interactions or relations between distinct biological components is called a biological network (e.g., genes, proteins, taxa, and metabolites). Now, with the availability of large-scale multi-omics data, the biological system has expanded from basic to advanced levels (like PPI connectivity and functional changes in disease stages, Disease-specific interactome, genetic perturbations, and network dysfunction). Protein–protein interaction networks (PINs), in which proteins act as nodes and interactions as undirected edges, depict the physical connections between the proteins that are present in a cell. Protein–protein interactions (PPIs) are the most thoroughly studied networks in biology and are crucial to cellular functions. PPIs can be found using a variety of experimental methods, with the yeast two-hybrid system being one of the more popular methods for studying binary interactions. Mass
5 Network Medicine: Methods and Applications
79
spectrometry-based high-throughput research have recently uncovered numerous sets of protein interactions. The databases that catalog experimentally determined protein–protein interactions have been the result of numerous international efforts over the past few decades, such as the Munich Information Center for Protein Sequence (MIPS) protein interaction database (Schoof et al. 2005), Biomolecular Interaction Network Database (BIND) (Bader 2003), Database of Interacting Proteins (DIP) (Xenarios 2000), Molecular Interaction database (MINT) (Chatr-aryamontri et al. 2007), and Protein Interaction database (IntAct) (Hermjakob 2004). These databases are classified into primary and secondary databases based on their interaction prediction method but now one new term introduced “Meta-database” which is a combination of different primary and secondary databases to get new and maximum protein–protein interactions in network (Fig. 5.2). The list of protein interaction databases is given in Table 5.1. Moreover PPI, the Metabolic networks explain the associations between small biomolecules (metabolites) and the enzymes (proteins) that interact with them to catalyze a biochemical reaction. Genetic interaction networks are valuable for understanding the relationship between genotype and phenotype because they relate to the functional interactions between pairs of genes in an organism. Similarly, a gene regulatory network (GRN) is a collection of molecular regulators that interact with one another and with other components of the cell to regulate the expression level of mRNA and protein. Further, when different cell signaling pathways interact, cellular signaling networks are built, and they are identified by a combination of experimental and computational techniques. A complex biological network consists of various nodes (including genes, proteins, and metabolites.) and connections among nodes are represented by joining lines, called “edges.” The hubs are nodes that connect with other nodes frequently and play important roles in biological processes. The total number of edges that a node is connected to is referred to as its degree. Finding the node with the top degree can support to distinguish a biological-entity that plays the most important role within the network. For more details about the basic biological network and its
Fig. 5.2 An example of human disease networks
80
A. Alam et al.
Table 5.1 Protein–protein interaction databases Primary database Databases BioGrid
URL https://thebiogrid.org/
Experimental/ Predicted Exp.
HPRD IntAct MINT
http://www.hprd.org/ https://www.ebi.ac.uk/intact/home https://mint.bio.uniroma2.it/
Exp. Exp. Exp.
HuRI http://www.interactome-atlas.org/ Secondary database STRING https://string-db.org/
Exp.
Reference Oughtred et al. (2019) Goel et al. (2012) Hermjakob (2004) Chatr-aryamontri et al. (2007) Luck et al. (2020)
Exp. & Pred.
Szklarczyk et al. (2019) UniHI http://www.unihi.org/ Exp. & Pred. Kalathur et al. (2014) Mentha http://mentha.uniroma2.it/ Exp. Calderone et al. (2013) APID http://cicblade.dep.usal.es:8080/ Exp. Alonso-López et al. APID/init.action (2019) HIPPIE http://cbdm-01.zdv.uni-mainz. Exp. Alanis-Lobato et al. de/~mschaefer/hippie/index.php (2017) HitPredict http://www.hitpredict.org/ Exp. Patil et al. (2011) IID http://iid.ophid.utoronto.ca/ Exp. & Pred. Kotlyar et al. (2016) HINT http://hint.yulab.org/ Exp. Das and Yu (2012) GPS-Prot http://gpsprot.org/ Exp. Fahey et al. (2011) • Primary protein interaction databases containing literature-curated PPIs for human proteins. • Secondary protein interaction databases containing predicted and curated from primary databases.
topological properties, importance can be read in our previous articles (Alam et al. 2019; Alam et al. 2021).
5.2.4 Human Disease Networks A disease gene network involves the incorporation of genes linked to specific disease phenotypes into an established human interactome. The hypothesis posits that genes and their corresponding products associated with a given disease have the propensity to interact and form clusters within a localized sub-network, as opposed to being randomly distributed across the entire human interactome (Lee and Loscalzo 2019). The unbiased analysis of pathobiological linkages between various disease processes has also been made possible by disease networks. To construct the human disease network, there are many ways including literature survey to find the disease-associated genes and construct network. Further, there are many good, comprehensive databases available DisGeNet, an extensive and carefully curated database, integrates data from multiple publicly accessible databases, such as “UniProt/
5 Network Medicine: Methods and Applications
81
SwissProt,” “Cancer Genome Interpreter (CGI),” “Comparative Toxicogenomic DatabaseTM (CTDTM),” “Orphanet,” “Mouse Genome Database (MGD),” “PsyGeNET,” “Genomics England,” “ClinGen,” and “Rat Genome Database (RGD).” These databases were utilized to collect information on genes associated with various diseases, which were subsequently compiled and organized within the DisGeNet database (Pinero et al. 2015). Recently, we analyzed three diseases (including Cardiovascular disease, Diabetes type-2, and Parathyroid adenoma) and found that 33 genes are common in these diseases (Fig. 5.2). Similarly, in our one more previous articles (Alam et al. 2022) where we identified 33 high-scoring significant modules that hub genes that are shared between tuberculosis (TB) and overlapping non-communicable diseases (NCDs) (lung cancer, rheumatoid arthritis, diabetes mellitus, Parkinson’s disease, and cardiovascular disease).
5.2.5 Drug–Target Interactions The conventional “one disease–one target reductionist approach” is not widely applicable since the majority of drugs exert their effects by targeting multiple proteins, resulting in a net pharmacological impact that encompasses both therapeutic and adverse effects. Network techniques provide valuable tools for predicting drug actions in complex biological contexts. In this context, drugs can be mapped onto the human interactome through their identified targets, where the effects of drugs on specific nodes or targets are represented by edges in the biological network. In our previous study, when the common disease genes related to TB and non-communicable disease (NCDs) that include lung cancer, rheumatoid arthritis, diabetes mellitus, Parkinson’s disease, and cardiovascular disease; and built a bipartite network (drugs and targets) by mapping the hub genes of the modules to their corresponding drugs using the DGIdb database, the results revealed that a significant portion of the target genes had multiple hits (as shown in Fig. 5.3). This indicates that genes interacting with a greater variety of drugs may be more intricately connected to the underlying mechanisms driving the pathological phenotype associated with these drugs. The presence of multiple drug interactions with a gene suggests its involvement in multiple pathways or biological processes relevant to the disease phenotype, emphasizing its potential significance in the context of therapeutic interventions and understanding disease mechanisms (Alam et al. 2022).
5.2.6 Drug–Drug Interaction A change in a drug’s impact on the body when it is combined with another drug. The absorption of either drug can be sped up, slowed down, or improved by a drug–drug interaction. This could alter the way one or both drugs work, increase or decrease their effects, or have negative consequences. Studies have demonstrated that a substantial proportion, ranging from approximately 37–60%, of hospitalized patients may have one or more potentially interacting drug combinations upon admission
CRP
FN1
SELE
EPIRUBICIN
NATALIZUMAB
ALICAFORSEN
BI-505
HYALURONIC ACID
RIFAMPICIN
IODIPAMIDE
RALTITREXED
EBSELEN
SODIUM LAURYL SULFATE
CEFOTAXIME
GADOBENATE DIMEGLUMINE
BISMUTH SUBSALICYLATE
IBUPROFEN LYSINE
FLURBIPROFEN
DERSALAZINE
NAPROXCINOD
ARUNDIC ACID
CARPROFEN
FENOPROFEN
GINSENG, ASIAN
PIROXICAM
BALSALAZID
MAGNESIUM SALICYLATE
CHEMBL374385
RESERPINE
AMINOSALICYLATE POTASSIUM
MELOXICAM
OXYPHENBUTAZONE
CAPECITABINE
LICOFELONE
ICOSAPENT ETHYL
NIMESULIDE
TOLMETIN SODIUM
MECLOFENAMATE SODIUM
ANTIPYRINE
OXAPROZIN POTASSIUM
BROMFENAC
PARECOXIB
SALICYLIC ACID
DIFLUNISAL
SULINDAC
CIMICOXIB
MECLOFENAMIC ACID
DIHOMO-GAMMA-LINOLENIC ACID
SALSALATE
DICLOFENAC SODIUM
DICLOFENAC POTASSIUM
INDOMETHACIN
ETORICOXIB
AMINOSALICYLATE SODIUM
LENALIDOMIDE
VALDECOXIB
ICOSAPENT
AMINOSALICYLIC ACID
APRICOXIB
NAPROXEN ETEMESIL
TENOXICAM
MEFENAMIC ACID
ANTRAFENINE
POLMACOXIB
DICLOFENAC
CHEMBL364804
LUMIRACOXIB
KETOROLAC
FENOPROFEN CALCIUM
SUPROFEN
NAPROXEN SODIUM
BALSALAZIDE DISODIUM
NABUMETONE
OXAPROZIN
POMALIDOMIDE
OLSALAZINE SODIUM
PHENTERMINE HYDROCHLORIDE
PHENYLBUTAZONE
KETOPROFEN
DICLOFENAC EPOLAMINE
SULFASALAZINE
FLUFENAMIC ACID
NIFLUMIC ACID
PARECOXIB SODIUM
TIAPROFENIC ACID
LORNOXICAM
NAPROXEN
CHEMBL68211
IBUPROFEN SODIUM
NEPAFENAC
FIROCOXIB
ETODOLAC
KETOROLAC TROMETHAMINE
INDOMETHACIN SODIUM
FLURBIPROFEN SODIUM
BROMFENAC SODIUM
TOLMETIN
HYDROXYCHLOROQUINE
RA-18C3
MITOMYCIN
RALOXIFENE
RILONACEPT
AMG-108
DIACEREIN
GALLIUM NITRATE
ACITRETIN
MAFOSFAMIDE
THYROGLOBULIN
CEFACLOR
BELNACASAN
NIMUSTINE HYDROCHLORIDE
HYDROCORTISONE
CLODRONATE DISODIUM
ETIPREDNOL DICLOACETATE
NICARDIPINE
CANAKINUMAB
PENTAMIDINE
GEVOKIZUMAB
ERYTHROMYCIN SODIUM beta-NICOTINAMIDE ADENINE DINUCLEOTIDE PHOSPHATE CELASTROL
PROXYPHYLLINE
MANNITOL
IXEKIZUMAB
ASPIRIN
SECUKINUMAB
TETRACYCLINE
METHOTREXATE
OFLOXACIN
ROFECOXIB
CHEMBL197194
GENTAMICIN
GARLIC
AZIRIDINE
ACETAMINOPHEN
METOCLOPRAMIDE
PENTOXIFYLLINE
TRICHOSTATIN
GIRENTUXIMAB
CEFEPIME
TROPISETRON
HYDROQUINONE
LANSOPRAZOLE
PRASTERONE
VERAPAMIL
DIDEMNIN B
CHEMBL429852
PENTAMIDINE ISETHIONATE
BECLOMETHASONE
CALCITRIOL
MIDOSTAURIN
PYRIMETHAMINE
QUININE
CLOZAPINE
IMIPENEM
INTERFERON BETA-1B
AMIKACIN SULFATE
PROCHLORPERAZINE
TARGININE
ROXITHROMYCIN
DIETHYLSTILBESTROL DIPHOSPHATE
MESALAMINE
GLYBURIDE
BROMOCRIPTINE
CHEMBL106187
IBUDILAST
FENTANYL
ALPRAZOLAM
TOCOPHEROL ACETATE
DIPYRIDAMOLE
RIVANICLINE
ALUMINUM HYDROXIDE
CHEMBL262489
TALC
CIDOFOVIR
THALIDOMIDE
PAMIDRONIC ACID
METHYLENE BLUE
FOSCARNET
RUTIN
CEFTRIAXONE
ABX-IL8
LEFLUNOMIDE
FENRETINIDE
MIDAZOLAM
CETUXIMAB
MEDROXYPROGESTERONE ACETATE
IBUPROFEN
HuMax-IL8
TYLOSIN
EDARAVONE
IONOMYCIN
CHEMBL1213492
MELATONIN
CLARITHROMYCIN
RABEPRAZOLE
BECLOMETHASONE DIPROPIONATE
CYTARABINE
SCH-708980
AMOXICILLIN
RESVERATROL
ACYCLOVIR
LISOFYLLINE
SULINDAC SULFIDE
FLUTICASONE PROPIONATE
SIROLIMUS
PEFLOXACIN
SIMVASTATIN
MEGESTROL ACETATE
APREMILAST
VX-702
CYCLOPHOSPHAMIDE
OXALIPLATIN
TALMAPIMOD
TRETINOIN
TICARCILLIN
AMG-714
ATPGAMMAS
HALOFUGINONE
PRAVASTATIN
OMEPRAZOLE
LITHIUM
CYCLOSPORINE
WORTMANNIN
PHORBOL MYRISTATE ACETATE
ATEZOLIZUMAB
IPILIMUMAB
METHIMAZOLE
TREMELIMUMAB
BLEOMYCIN (CHEMBL3039590)
TERFENADINE
PREDNISONE
MELPHALAN
INTERLEUKIN 1BETA
AMITRIPTYLINE
FONTOLIZUMAB
TRASTUZUMAB
FUMARIC ACID
MYCOPHENOLATE MOFETIL
THEOPHYLLINE
OLSALAZINE
SURAMIN
THROMBIN
GANCICLOVIR
INTERFERON ALFA-2B
AMIKACIN
NORDIHYDROGUAIARETIC ACID
VINORELBIN DITARTRATE
ROQUINIMEX
ETANERCEPT
PERFOSFAMIDE
MILTEFOSINE
SODIUM SALICYLATE
TEMOZOLOMIDE
CILOSTAZOL
TRIFLUOPERAZINE
MOR-103
FAMCICLOVIR
MECHLORETHAMINE
DITIOCARB
LACTULOSE HYDRATE
KB-003
IDARUBICIN
STAUROSPORINE
NAMILUMAB
ISOPROTERENOL
CETIRIZINE
BCG VACCINE
TETANUS TOXOID
PROCARBAZINE
ACLARUBICIN
MT-203
DIAZIQUONE
INDINAVIR
GLUCOSAMINE
VINBLASTINE
STREPTOZOTOCIN
MEDRONIC ACID
CHONDROITIN SULFATE
BEVACIZUMAB
PACLITAXEL
CELECOXIB
RIZATRIPTAN
ZOLEDRONIC ACID
LEVODOPA
ALCOHOL
CHLORHEXIDINE
LETROZOLE
CHEMBL1233506
BATIMASTAT
CHEMBL181244
REBIMASTAT
RAMIPRIL
FILGRASTIM
TANOMASTAT
ETRETINATE
SOLIMASTAT
TILUDRONIC ACID
ILOMASTAT
MARIMASTAT
INCYCLINIDE
CAPTOPRIL
ITRACONAZOLE
CGS-27023A
PRINOMASTAT
TROGLITAZONE
URIDINE DIPHOSPHATE GLUCOSE
METHYLDOPA
NIFEDIPINE
ANDECALIXIMAB
DOXIFLURIDINE
METHYLPREDNISOLONE
IDRONOXIL
EMRICASAN
CHEMBL586058
CHEMBL310554
CHEMBL293034
CHEMBL417149
RALIMETINIB
SCIO-323
CHEMBL380373
URSODIOL
RWJ-67657
FX-005
DOXYCYCLINE
CHEMBL193264
CHEMBL522579
VX-745
CHEMBL379760
PH-797804
CHEMBL1236539
Acumapimod
CHEMBL29641
MORPHINE
CHEMBL94417
TA-5493
CHEMBL273158
CHEMBL258202
SC-80036
BMS-582949
CHEMBL425616
CHEMBL252128
2-METHOXYESTRADIOL
CHEMBL96741
Ro-3201195
CHEMBL87277
SB-85635
CHEMBL492265
HYDROGEN PEROXIDE
CHEMBL1951415
SB-202190
CHEMBL85860
CHEMBL590753
RALIMETINIB MESYLATE
AZD-7624
LOSMAPIMOD
GSK-610677
AVE-9940
ARRY-797
PG-760564
CHEMBL237127
DILMAPIMOD
CHEMBL410668
KC-706
SD-0006
DORAMAPIMOD
Pexmetinib
CHEMBL199237
AZD-6703
RAVIDASVIR
CHEMBL197277
CHEMBL313417
PIRFENIDONE
PS-516895
CHEMBL564912
PROPOXYPHENE
2-CHLOROPHENOL
LY-3007113
CHEMBL257341
AMG-548
CHEMBL478649
METHOXSALEN
TAK-715
CHEMBL365776
LEO-15520
VASOPRESSIN
CHEMBL191598
PAMAPIMOD
SEMAPIMOD
CHEMBL494072
CHEMBL559401
CHEMBL193156
PF-03715455
CHEMBL371491
GW607117X
CHEMBL194009
CHEMBL305178
R-1487 (CHEMBL1766582)
CHEMBL194322
CHEMBL195393
HALOPERIDOL
CHEMBL195450
CHEMBL522387
EPOETIN BETA
CHEMBL306380
DOXORUBICIN
SB-203580
SB-220025
Refametinib
MK-8353
CHEMBL369507
IROFULVEN
ARSENIC TRIOXIDE
CHEMBL387385
BENZOQUINONE
ERLOTINIB
URACIL
CHIR-99021
Ulixertinib
CHEMBL259551
ETHER
ISOPRENALINE
REBAMIPIDE
GDC-0994
HYPOTHEMYCIN
CHEMBL1229592
GONADOTROPIN, CHORIONIC
REGRAMOSTIM
BUMETANIDE
OLOMOUCINE
CHEMBL219841
TURPENTINE
STAVUDINE
ZIDOVUDINE
ATORVASTATIN
NI-0801
ELDELUMAB
ANTIBIOTIC
TESTOSTERONE
ATROPINE
RITONAVIR
NELFINAVIR
MYRISTIC ACID
INFLIXIMAB
DODECANOATE
EPIGALOCATECHIN GALLATE
SAQUINAVIR
ERITORAN (CHEMBL501259)
FOLIC ACID
CHEMBL225157
ERITORAN TETRASODIUM (CHEMBL3301672)
CHEMBL1232858
GOLOTIMOD
CHLORAMPHENICOL
VOLANESORSEN SODIUM
SULFURIC ACID
UREA
CHEMBL173373
HEXAMETHYLENEBISACETAMIDE
QUINACRINE DIHYDROCHLORIDE
AMPICILLIN
SODIUM CHLORIDE
MERCAPTOPURINE
SUCCINOBUCOL
DEXAMETHASONE
MYCOPHENOLIC ACID
TACROLIMUS
THYROXINE
COLCHICINE
DEXTRAN SULFATE SODIUM
DANAZOL
CHEMBL134074
CARLUMAB
PREDNISOLONE
CARVEDILOL
PUROMYCIN
QUERCETIN
BIMOSIAMOSE
CHEMBL1234621
RIVIPANSEL
OCRIPLASMIN
L19SIP
L19IL2
L19TNFa
AS-1409
PASCOLIZUMAB
ANAKINRA
ROLOFYLLINE (CHEMBL52333)
PHENPROCOUMON
METHIONINE SULFOXIDE
Burixafor
CHEMBL452864
PLERIXAFOR
BKT140
MSX-122
POL6326
CTCE-9908
CHEMBL1088913
FRAMYCETIN
CHEMBL460491
ULOCUPLUMAB
CHEMBL518924
CHEMBL406845
PHOSPHOCHOLINE
ENLIMOMAB PEGOL
LIFITEGRAST
TGFB1
SPP1
LEP
MMP1
PPARG
EGFR
IGF1
SERPINE1
CHEMBL1235854
DUSIGITUMAB
DALTEPARIN SODIUM
DROTRECOGIN ALFA (ACTIVATED)
REGRAMOSTIM
UROKINASE
FIBRINOLYSIN, HUMAN
ARSENIC TRIOXIDE
NIMODIPINE
VASOPRESSIN
ESTROGENS, CONJUGATED
EPIRUBICIN
DEFIBROTIDE
FLUOXETINE
RETEPLASE
ALTEPLASE
CAPTOPRIL
GLYCYRRHIZIN
ORLISTAT
HYDROCHLOROTHIAZIDE
CETRORELIX
URIDINE
TENECTEPLASE
LEVOTHYROXINE
ALEPLASININ
ANISTREPLASE
SAPITINIB
DASATINIB
TESEVATINIB
S-222611
AMG-337
ZALUTUMUMAB
NEPIDERMIN
SUNITINIB
CAPREOMYCINE
MP-412
CANERTINIB DIHYDROCHLORIDE
CHEMBL387187
LIDOCAINE
IBRUTINIB
BUPARLISIB
CEP-32496
PROCATEROL
LINSITINIB
Futuximab
MK-2206
GOLVATINIB
IMGATUZUMAB
ROCILETINIB
CRIZOTINIB
OSIMERTINIB MESYLATE
NIMOTUZUMAB
CHEMBL306380
CETUXIMAB
BMS-754807
IGN-311
NERATINIB
PEMBROLIZUMAB
NVP-AUY922
AEE-788
FALNIDAMOL
VANDETANIB
CUDC-101
PELITINIB
AMLEXANOX
REGORAFENIB
CANERTINIB
GELDANAMYCIN
ALVOCIDIB
DEPATUXIZUMAB
IRINOTECAN
SARACATINIB
PF-00477736
DEPATUXIZUMAB MAFODOTIN
GEDATOLISIB
mAb-425
PEMETREXED (CHEMBL1201258)
DULIGOTUZUMAB
CHEMBL1229592
EGF816
ERISMODEGIB
AFATINIB DIMALEATE
VEMURAFENIB
TEMSIROLIMUS
MDX-447
CHEMBL1081312
MGCD-265
MUBRITINIB
ATEZOLIZUMAB
DABRAFENIB
Pyrotinib
OSIMERTINIB
CHEMBL1765740
MATUZUMAB
NIVOLUMAB
VARLITINIB
TYRPHOSTIN AG-1478
Savolitinib
AC-480
FELYPRESSIN
MOMELOTINIB
PACLITAXEL
HM-61713
CAPMATINIB
CHEMBL1231206
Poziotinib
Puquitinib
JNJ-26483327
CHEMBL1951415
NITROGLYCERIN
PANITUMUMAB
AZD-4547
LAPATINIB DITOSYLATE
ACALABRUTINIB
BOSUTINIB
INSM-18
SORAFENIB
CHEMBL174426
PHA-665752
POTASSIUM CHLORIDE
OLMUTINIB
JNJ-42756493
CARBOPLATIN
ERLOTINIB
PD-0166285 (CHEMBL3545196)
AZD-4769
TEMOZOLOMIDE
CABOZANTINIB
CHEMBL458997
TRASTUZUMAB
PONATINIB
DOVITINIB
TAK-285
BGB-283
BRIGATINIB
GEFITINIB
Allitinib
CHEMBL53753
ALCOHOL
TRAMETINIB
PIMASERTIB
PKI-166
Simotinib
BMS-690514
ERLOTINIB HYDROCHLORIDE
DACOMITINIB
LAPATINIB
ENCORAFENIB
IMATINIB
AFATINIB
Theliatinib
NAQUOTINIB
RG-7160
BEVACIZUMAB
NECITUMUMAB
EVEROLIMUS
ICOTINIB
GANETESPIB
DECITABINE
DACTOLISIB
CHEMBL285063
DURVALUMAB
SELUMETINIB
RINDOPEPIMUT
Epitinib
DACOMITINIB HYDRATE
SB-243213
CHEMBL2141478
CHEMBL56543
ORANTINIB
DICLOFENAC
ICOSAPENT
CHEMBL86658
CHEMBL200495
CHEMBL1946409
MBX-2044
OLSALAZINE SODIUM
GW501516
MURAGLITAZAR
OMS405
NETOGLITAZONE
GLIPIZIDE
INDEGLITAZAR
CHEMBL434063
CHEMBL364748
BEXAROTENE
BARDOXOLONE METHYL
ALEGLITAZAR
INT131
SODELGLITAZAR
CHEMBL191060
CHEMBL1230670
RIVOGLITAZONE
IMIGLITAZAR
CHEMBL24458
LY-510929
NAVEGLITAZAR
TAGITININ A
CHEMBL201880
REGLITAZAR
CHEMBL23296
CHEMBL410478
CHEMBL375270
INDOMETHACIN
BALSALAZIDE DISODIUM
EDAGLITAZONE
TIROTUNDIN
TELMISARTAN
BARDOXOLONE
NATEGLINIDE
GSK-9578
MITIGLINIDE
ETALOCIB
TESAGLITAZAR
CHEMBL1210221
CHEMBL510698
CHEMBL378160
Efatutazone hydrochloride
RESVERATROL
CHEMBL451721
CHEMBL191275
LINOLEIC ACID
BALAGLITAZONE
CHEMBL1204498
BEZAFIBRATE
RAGAGLITAZAR
PIOGLITAZONE HYDROCHLORIDE
TREPROSTINIL
EFATUTAZONE
ROSIGLITAZONE
ICOSAPENT ETHYL
MK-0533
DB959
CHEMBL149394
METAGLIDASEN
SULFASALAZINE
ELAFIBRANOR
CHEMBL88496
FARGLITAZAR
GW7845
ARHALOFENATE
CIGLITAZONE
L-796449
L-165461
ROSIGLITAZONE MALEATE
CHEMBL179330
CELECOXIB
CHEMBL146624
REPAGLINIDE
MESALAMINE
ATx08-001
CHEMBL181937
CLX-0921
AVE0847
BALSALAZID
SIROLIMUS
OLANZAPINE
ETOPOSIDE
PIOGLITAZONE
SOYBEAN OIL
GEMFIBROZIL
MIRTAZAPINE
VALPROIC ACID
METHIMAZOLE
CLOZAPINE
METHADONE
RISPERIDONE
LOVASTATIN
BROMOCRIPTINE
ESTRADIOL VALERATE
FELODIPINE
PROPYLTHIOURACIL
CARVEDILOL
GENISTEIN
ZOLEDRONIC ACID
FOLIC ACID
TROGLITAZONE
LAMIVUDINE
LEFLUNOMIDE
REBIMASTAT
PRINOMASTAT
RIBAVIRIN
CHEMBL11306
MARIMASTAT
DOXYCYCLINE HYDRATE
COLLAGENASE CLOSTRIDIUM HISTOLYTICUM
CGS-27023A
DOXYCYCLINE HYCLATE
INTERFERON BETA
PENTOSAN POLYSULFATE SODIUM
CHEMBL92608
CIPEMASTAT
MEDROXYPROGESTERONE ACETATE
SOLIMASTAT
ILOMASTAT
DOXYCYCLINE CALCIUM
BATIMASTAT
CHEMBL45483
LEUPROLIDE ACETATE
TRIAMCINOLONE
VITAMIN E
HYDROCORTISONE
ESTRADIOL
TAMOXIFEN
ISOTRETINOIN
METELIMUMAB
IDOXURIDINE
HYALURONIDASE (HUMAN RECOMBINANT)
CLADRIBINE
STREPTOZOTOCIN
DOXORUBICIN
ATENOLOL
INOSITOL
ISOFLAVONE
INTERFERON ALFA-2B
GLATIRAMER ACETATE
GOSSYPOL
GALLIUM NITRATE
RAMIPRIL
TESTOSTERONE
PIRFENIDONE
AMIFOSTINE
LY-2382770
VERAPAMIL
FENRETINIDE
EMODIN
TOREMIFENE
MELATONIN
NICOTINE
HYDROGEN PEROXIDE
FRESOLIMUMAB
TRETINOIN
PHORBOL MYRISTATE ACETATE
SODIUM BUTYRATE
WORTMANNIN
GENTAMICIN
ASK-8007
TACROLIMUS
CHONDROITIN SULFATE
CALCITONIN
ETRETINATE
STAT3
SIRT1
VEGFA
CD44
CDKN1A
MAPK8
ESR1
HGF
PLG
DOCETAXEL
ETOPOSIDE
CHEMBL89363
MMP13
QUIZARTINIB
REBAMIPIDE
NYSTATIN
ALTEPLASE
UROKINASE
AMINOCAPROIC ACID
PREDNISOLONE
FIBRINOLYSIN, HUMAN
INOSITOL
COUMARIN
ANISTREPLASE
CHEMBL2016873
SOYBEAN OIL
CHEMBL35482
DEFIBROTIDE SODIUM
TENECTEPLASE
LEPIRUDIN
GARLIC
APROTININ
PHORBOL MYRISTATE ACETATE
AMEDIPLASE
PENICILLIN G SODIUM
DESMOTEPLASE
NORETHINDRONE
TRANEXAMIC ACID
STREPTOKINASE
OXYMETHOLONE
RETEPLASE
DANAZOL
GENISTEIN
2-AMINO-1-METHYL-6- PHENYLIMIDAZO[4,5B]PYRIDINE
TOREMIFENE
ESTRADIOL ACETATE (CHEMBL1200430)
CHEMBL391910
DIETHYLSTILBESTROL DIPHOSPHATE
GONADORELIN
PERTUZUMAB
DROLOXIFENE
OSPEMIFENE
CHEMBL188528
CHEMBL201013
MEGESTROL ACETATE
ESTRONE SODIUM SULFATE
CHEMBL304552
CHEMBL282489
ALLYLESTRENOL
ESTROGENS, CONJUGATED SYNTHETIC A
LAPATINIB
MITOTANE
RALOXIFEN
TAMOXIFEN
LY2245461
BAZEDOXIFENE ACETATE
ESTRADIOL VALERATE
TAMOXIFEN CITRATE
ENDOXIFEN
NORGESTREL
ESTROGENS, ESTERIFIED
QUINESTROL
LASOFOXIFENE
IDOXIFENE
ARZOXIFENE
CUSTIRSEN
ERTEBEREL
CHEMBL1213270
ESTRONE
TOREMIFENE CITRATE
TIBOLONE
CHEMBL184151
CARBOQUONE
ESTRIOL
AFIMOXIFENE (CHEMBL10041)
CHEMBL222501
SIVIFENE
CLOMIPHENE
DESOGESTREL
EVEROLIMUS
FLUOXYMESTERONE
PRASTERONE
LETROZOLE
CHEMBL180300
PRINABEREL
FULVESTRANT
LEVONORGESTREL
AFIMOXIFENE (CHEMBL489)
MK-6913
MESTRANOL
ESTROPIPATE
TRILOSTANE
GDC-0810
SR16234 (CHEMBL3545210)
ETHINYL ESTRADIOL
RALOXIFENE
RALOXIFENE CORE
ETHYNODIOL DIACETATE
ESTRADIOL CYPIONATE
DIENESTROL
DIENOGEST
NORGESTIMATE
POLYESTRADIOL PHOSPHATE
ACOLBIFENE
ETONOGESTREL
ESTROGENS, CONJUGATED
SYNTHETIC CONJUGATED ESTROGENS, B
PROPYLPYRAZOLETRIOL
CHEMBL236086
PALBOCICLIB
CHEMBL181936
DIARYLPROPIONITRILE
VINTAFOLIDE
PROGESTERONE
GTx-758
MELATONIN
CHEMBL180517
CHF4227
IODINE
CHEMBL180071
CHEMBL193676
FISPEMIFENE
ESTRADIOL
EXEMESTANE
LEFLUNOMIDE
ANASTROZOLE
ESTRAMUSTINE
CHLOROTRIANISENE
NORELGESTROMIN
ABEMACICLIB
HEXESTROL
CHEMBL223026
BAZEDOXIFENE
ESTROGEN
CLOMIPHENE CITRATE
DIETHYLSTILBESTROL
Ribociclib
MEDROXYPROGESTERONE ACETATE
CHEMBL520107
CHEMBL236718
RAD1901
TRASTUZUMAB
RILOTUMUMAB
CABOZANTINIB
WORTMANNIN
GONADOTROPIN, CHORIONIC
FICLATUZUMAB
SODIUM HEPARIN
THALIDOMIDE
FORETINIB
ASPIRIN
EPIGALOCATECHIN GALLATE
RESERPINE
METHOTREXATE
STREPTOZOTOCIN
RESVERATROL
CHEMBL262489
SPLITOMICIN
CHEMBL257991
CHEMBL420311
SODIUM LAURYL SULFATE
CARVEDILOL
BEVACIZUMAB
CILOSTAZOL
PEGAPTANIB SODIUM
TROMETHAMINE
VANDETANIB
FENOFIBRATE
MUPARFOSTAT (CHEMBL1615835)
DALTEPARIN SODIUM
CONBERCEPT
SEMAXANIB
CHEMBL88153
BEVASIRANIB
PEGAPTANIB OCTASODIUM
PENTOSAN POLYSULFATE SODIUM
BROLUCIZUMAB
BEVACIZUMAB 111IN
GLICLAZIDE
LENALIDOMIDE
CHEMBL313417
PHENYTOIN
RANIBIZUMAB
BEVASIRANIB SODIUM
PYROGLUTAMIC ACID
SORAFENIB
LITHOCHOLIC ACID
AFLIBERCEPT
TAK-593
JUGLONE
TNF
CXCL12
CASP9
DECR1
CAT
ASPIRIN
PYRAZOLE
SORAFENIB TOSYLATE
GENTAMICIN
PROPYLTHIOURACIL
NAFAMOSTAT
FLUOCINOLONE ACETONIDE
PROCARBAZINE
POMALIDOMIDE
CROMOLYN SODIUM
MIDAZOLAM
CERTOLIZUMAB PEGOL
NORDIHYDROGUAIARETIC ACID
CELASTROL
GOLIMUMAB
GLIMEPIRIDE
CARBAMAZEPINE
INAMRINONE
GENTAMICIN
PHORBOL MYRISTATE ACETATE
THALIDOMIDE
CEFOTAXIME
DERSALAZINE
INFLIXIMAB
PENICILLIN G SODIUM
BUPIVACAINE
ADALIMUMAB
PYRIDOXINE
GLUCOSAMINE
ETANERCEPT
PRAZOSIN HYDROCHLORIDE
OMEPRAZOLE
RUTIN
LACTULOSE HYDRATE
IBUDILAST
APREMILAST
MILTEFOSINE
RISPERIDONE
MIDOSTAURIN
CHLOROQUINE
RABEPRAZOLE
AZ-9773
ABACAVIR
PEGSUNERCEPT
CHEMBL173373
OZORALIZUMAB
MAGNESIUM SULFATE
ONERCEPT
PRANLUKAST
AFELIMOMAB
HYDROXYCHLOROQUINE
NERELIMOMAB
TALMAPIMOD
LENERCEPT
VX-702
ORTATAXEL
CLENBUTEROL
LENALIDOMIDE
BETA-LAPACHONE
PENTOXIFYLLINE
PF-04236921
AJULEMIC ACID
METHIMAZOLE
PIRFENIDONE
SPIRONOLACTONE
VADIMEZAN
FOLIC ACID
CHEMBL219629
DIDANOSINE
DELMITIDE
HALOFUGINONE
PLACULUMAB
MEROPENEM
URAPIDIL
TIMOLOL MALEATE
TALACTOFERRIN ALFA
TINZAPARIN SODIUM
NIVOCASAN
EMRICASAN
CELECOXIB
CHEMBL174539
CHEMBL1161862
FOMEPIZOLE
DOXYCYCLINE HYDRATE
CHEMBL496942
3-METHYLPYRIDINE
DOXYCYCLINE CALCIUM
MARIMASTAT
CHEMBL116757
CHEMBL222002
PRINOMASTAT
URIDINE DIPHOSPHATE GLUCOSE
CHEMBL327002
DOXYCYCLINE HYCLATE
CHEMBL236573
CHEMBL514794
CHEMBL468900
CHEMBL511942
CHEMBL409536
CHEMBL1234479
CHEMBL93687
CHEMBL70176
CHEMBL440498
PHENYLEPHRINE
ATORVASTATIN
ESTRADIOL DIPROPIONATE
INTERFERON GAMA-1B
HYALURONIC ACID
MOMETASONE FUROATE
MYCOPHENOLIC ACID
CELECOXIB
STAUROSPORINE
FAMOTIDINE
Sucralfate
PENTOSAN POLYSULFATE SODIUM
MUPARFOSTAT (CHEMBL1615835)
CHEMBL262489
TRIAMCINOLONE
CAFFEINE
VINCRISTINE
MEDRONIC ACID
DEPSIPEPTIDE
CYCLOSPORINE
CLADRIBINE
NIMUSTINE HYDROCHLORIDE
VINBLASTINE
MELPHALAN
TIPIFARNIB
ACETAMINOPHEN
DEOXYCYTIDINE
PACLITAXEL
DICUMAROL
FLUTAMIDE
CARBOPLATIN
CHEMBL3186227
SODIUM PHENYLBUTYRATE
NICOTINE
ARSENIC TRIOXIDE
IRINOTECAN
EPOETIN BETA
SIROLIMUS
INDOMETHACIN
FULVESTRANT
THYROTROPIN
DUSIGITUMAB
SODIUM SALICYLATE
FGF2
GLUTAMINE
CABAZITAXEL
BETAMETHASONE
EDARAVONE
DIAZIQUONE
FLUOROURACIL
DEXTRAN SULFATE SODIUM
HEXANE
VITAMIN E
SP-600125
CHEMBL210618
BLEOMYCIN (CHEMBL3039590)
GEFITINIB
CHEMBL382639
BUTYLATED HYDROXYTOLUENE
TRETINOIN
CHEMBL209740
Bentamapimod
ENALAPRIL
OXYTOCIN
2-METHOXYESTRADIOL
CHEMBL242237
BEXAROTENE
ANASTROZOLE
GEFITINIB
LINSITINIB
ANTAZOLINE HYDROCHLORIDE
FLUOXETINE
IGF2
CHEMBL207228
PRAZOSIN
DOXYCYCLINE
DEXRAZOXANE
CC-401
TANZISERTIB
METARAMINOL
ACITRETIN
BARDOXOLONE METHYL
ATIPRIMOD
APOA1
APOE
CST3
IL6
IL2RA
IL22
ALCOHOL
FENOFIBRATE
LAMIVUDINE
GLUCAGON
TESTOSTERONE
APABETALONE
FUROSEMIDE
RITONAVIR
TRETINOIN
ALBUMIN HUMAN
STAUROSPORINE
GINKGO
IRBESARTAN
VITAMIN E
SIMVASTATIN
LORAZEPAM
PREDNISONE
LUTEIN
BUTHIONINE SULFOXIMINE
SOYBEAN OIL
GONADOTROPIN, CHORIONIC
TRIAMCINOLONE
TROGLITAZONE
GANCICLOVIR
CERIVASTATIN
RIBAVIRIN
DIGOXIN
SAQUINAVIR
FENTANYL
CDP-6038
CLAZAKIZUMAB
IBUDILAST
PF-04236921
ELSILIMOMAB
OLOKIZUMAB
SIRUKUMAB
SILTUXIMAB
GINSENG, ASIAN
ARSENIC TRIOXIDE
METRONIDAZOLE
VITAMIN K
LEVOFLOXACIN
VX-702
IFOSFAMIDE
INTERFERON ALFA-2B
LINEZOLID
GEMFIBROZIL
NELFINAVIR
GALLIUM NITRATE
DIHYDROSPINGOSINE
MIDOSTAURIN
DIDANOSINE
METHIMAZOLE
LENTINAN
CAFFEINE
METHYLPREDNISOLONE
GONADOTROPIN, CHORIONIC
MERCAPTOPURINE
PROPYLTHIOURACIL
VERAPAMIL
STAUROSPORINE
AS-1409
FLUOROURACIL
INTERFERON BETA-1B SODIUM beta-NICOTINAMIDE ADENINE DINUCLEOTIDE PHOSPHATE FLUTAMIDE
DECITABINE
THEOPHYLLINE
PROGESTERONE
BECLOMETHASONE DIPROPIONATE
MEGESTROL ACETATE
PACLITAXEL
CLADRIBINE
L-778123
CETIRIZINE
ATORVASTATIN
ISOTRETINOIN
LMB-2
THYROXINE
GARLIC
DEXTRAN SULFATE SODIUM
BRYOSTATIN
INOLIMOMAB
DENILEUKIN DIFTITOX
HUMAN INTERLEUKIN-2
BASILIXIMAB
DACLIZUMAB
FEZAKINUMAB
MPO
HMOX1
SOD2
CYCLOPHOSPHAMIDE ANTIBIOTIC CHEMBL1161866
CCR5
THEOPHYLLINE
TENECTEPLASE
NIMESULIDE
NOMIFENSINE
BROMOACETIC ACID
PSORALEN
PYRAZINAMIDE
DIMETHYL SULFOXIDE
FENTANYL
LIDOCAINE
FLUTAMIDE SODIUM beta-NICOTINAMIDE ADENINE DINUCLEOTIDE PHOSPHATE AMSALOG
METHYLENE BLUE
CEFDINIR
MELATONIN
CHEMBL310871
LORATADINE
TRIMETHOPRIM
TOLMETIN
DOXYCYCLINE
TOCOPHEROL ACETATE
MESALAMINE
DICLOFENAC
LEVOCARNITINE
FLUDARABINE
OCTREOTIDE
LITHIUM
FILGRASTIM
BILIVERDIN
FORMIC ACID
ASPIRIN
ZINC CHLORIDE
SUNITINIB
SORAFENIB
IL23A
CHEMBL1196395
CENICRIVIROC
AZD5672
APLAVIROC
ANCRIVIROC (CHEMBL2107311)
GLATIRAMER ACETATE
CHEMBL518924
ENOXAPARIN
CHEMBL41275
GLYCERIN
PRO-140
BICARBONATE
CHEMBL207004
GUANINE
CHEMBL78535
GUANIDINO ACETATE
VICRIVIROC MALEATE
APLAVIROC HYDROCHLORIDE
INCB-9471
PF-04634817
PHENPROCOUMON
VICRIVIROC
MARAVIROC
USTEKINUMAB
TILDRAKIZUMAB
GUSELKUMAB
BRIAKINUMAB
HP
LPA
HLA-DQB1
CETP
HLA-A
HLA-DRB1
HLA-DQA1
AMOXICILLIN
CLAVULANIC ACID
CERIVASTATIN
TORCETRAPIB
DALCETRAPIB
ANACETRAPIB
TAMOXIFEN
EVACETRAPIB
CHEMBL67129
INDOLE-3-BUTYRIC ACID
CARBAMAZEPINE
LYM-1
GLATIRAMER ACETATE
LAPATINIB
INTERFERON BETA-1A
INTERFERON BETA-1B
PYRIDOXINE
STREPTOZOTOCIN
ESTRADIOL
AMINOCAPROIC ACID
HLA-B
Module-13 od du d ullle u e
Module-10 ulle e
Mod M Mo odu od du ule lele le-9 e9 e-9
Module-7 Module Mo M Mod odu dule-7 dule ule-7
Module-6 M od du ule-6 e--6 6
Module-5 M d -5 5
OXCARBAZEPINE PHENYTOIN ABACAVIR THALIDOMIDE PAZOPANIB CARBAMAZEPINE
Fig. 5.3 Drug–target interaction network: all the 13 modules’ targets (Green) are mapped with their respective drugs (Magenta). The bar graph illustrating the number of interacting drugs with key targets (Ref: doi:10.3389/fphar.2021.770762)
ALB
PTGS2
IL1RN IL1A
IL1B
IL17A
IL2
CXCL8
IL10
CCL5
IL15
CTLA4
IFNG
CSF2
MMP2
MMP9
CASP3
MAPK14
MAPK1
CXCL10
TLR4
TLR2
HMGB1
VCAM1
IL18
CCL2
CD34
Module-2 Module Modul Modu odule-2 oduleodule odu du dule dule-2 duledu ullle le-2 lee-2 e e-
ICAM1
Module-3 M Mod Mo o le e--3 e e-3
NLRP3 IL4
Module-4 M Mod Mo o ule e-4 4
FOXP3
Mo Mod Modu M Module ModuleModule-1 odul od o oduleodu dule-1 d dule ule ule ul ule-1 llee -1
Number of interacting drugs
Module-11 Mod M Mo od o du ulle e-11
CXCR4
82 A. Alam et al.
5 Network Medicine: Methods and Applications
83
(Costa 1991). Digoxin, beta-blockers, estrogen, oral hypoglycemic medicines, and diuretics were the five drug classes most likely to be involved in possible drug interactions. In this way, Cao et al. have developed a database, e.g., DDInter (http:// ddinter.scbdd.com) (Xiong et al. 2022), The curated Drug–Drug Interaction (DDI) database offers extensive data, practical medication guidance, an intuitive functional interface, and powerful visualization tools to cater to the needs of the scientific community. The database currently includes approximately 0.24 million DDI associations. Figure 5.3 showcases the Drug–Target Interaction Network, where all 13 modules’ targets (in green) are mapped with their respective drugs (in magenta). The illustration on the right demonstrates the number of interacting drugs with key targets, the database connects 1833 approved drugs, encompassing 1972 entities. Each drug in the database is accompanied by essential chemical and pharmacological information, along with its interaction network. This comprehensive annotation allows for a deeper understanding of the drug’s characteristics and its interactions with other entities within the network (Fig. 5.4).
5.2.7 Functional Modules in Molecular Networks Protein clusters or sub-networks that exhibit dense connections are often regarded as potential functional modules within a given network. Another hypothesis proposes that proteins interacting with similar groups of other proteins in the network
Fig. 5.4 DDInter provides detailed annotations of each DDI association and enables users to conduct data query (image courtesy, please refer: https://doi.org/10.1093/nar/gkab880)
84
A. Alam et al.
Fig. 5.5 Guilt-by-association approach: different colors are used to denote proteins with known functions, while proteins with unknown functions are left uncolored. Transferring functional annotation from directly interacting proteins allows for the inference of protein function (indicated by arrows)
tend to have comparable functions. Even in cases where direct interactions are absent, these proteins contribute to similar biological functions and should be included in the same modules. However, both definitions of modules rely on a guilt- by-association approach, as illustrated in Fig. 5.5. For instance, if two proteins interact with each other, it is more likely that they share the same cellular functionalities compared to proteins that do not interact (Wang and Qian 2014). There are many tools for the identification of functional modules within the network, but few are widely used among the researcher including the LEV (leading eigenvector) method that detects the communities in network from package “igraph” in R (Newman 2006). Another approach called MCODE (Molecular Complex Detection) focuses on identifying densely connected regions within a network. MCODE aims to identify molecular complexes or clusters characterized by high connectivity and close proximity of nodes (Bader and Hogue 2003). Another is the DIAMOnD (DIseAse MOdule Detection), which we can investigate the local network neighborhood (LNN) near a particular set of known disease proteins, and this helps us identify potential new disease target protein (Ghiassian et al. 2015) and last one which I found very useful and accurate that is MTGO (Module detection via Topological information and Gene Ontology knowledge) is a method that identifies modules within a network using both network topology and the biological role of proteins based on Gene Ontology (GO) terms (Vella et al. 2018).
5.3 Drug–Repurposing Opportunities Network-based drug repurposing methodologies are based on the premise that drugs capable of interacting with multiple targets can demonstrate efficacy against specific diseases. Furthermore, these methodologies take into account the potential therapeutic implications when two drugs target the same protein. By leveraging the interconnected nature of biological networks, these approaches offer promising
5 Network Medicine: Methods and Applications
85
avenues for identifying new therapeutic uses for existing drugs. This approach proves to be effective in the discovery and development of drug molecules with new pharmacological or therapeutic indications. By targeting specific proteins through therapeutic combinations or drug repurposing, it becomes possible to improve clinical conditions in cases of comorbidity, enhance the potency of certain drugs, and achieve synergistic effects for better treatment outcomes (Imam et al. 2023).
5.4 Drug Side Effects Over a million significant injuries and fatalities are caused annually by adverse drug reactions (ADRs), which are very common. Currently, many machine learning and network-based approaches are used to predict the adverse drug. By leveraging features such as composition, structure, and binding affinity, researchers have employed methods that involve machine learning (ML) and deep learning techniques (Dara et al. 2022). As part of these ongoing efforts, a recent development is the T-ARDIS database (Galletti et al. 2021). T-ARDIS is a carefully curated compilation of relationships between proteins and Adverse Drug Reactions (ADRs). These associations are statistically evaluated and sourced from existing databases of drug–target and drug–ADR associations.
5.5 Integrating Omics Data with Networks: Challenges and Ways Multi-omics data will soon be regularly used in preclinical and clinical contexts, hence further constraints should apply standardized, rigorous bioinformatics techniques to process, normalize, and analyze the massive datasets from various study modalities. From this perspective, it should soon be possible to build networks of networks to determine how various biological aspects are interconnected. Beyond intracellular molecular networks, we are now combining the connections between various cell-types, organ-systems, hosts and microorganisms, hosts and environmental exposures, as well as other interconnections. It would also be essential to incorporate psychosocial components when defining disease networks in order to link all contributing factors to clinical results. As demonstrated, network controllability analysis continues to be useful for identifying important disease genes and prioritizing potential targets. The existing human interactome is incomplete, which restricts network analysis of complicated human disorders. The interactome is expected to keep growing as high-throughput technology and bioinformatics continue to progress. The existing interactome is based solely on the binary biophysical interactions between the curated PPIs, with little knowledge of or annotations for protein-binding patterns or domains. There is an ongoing attempt to develop a “domain-specific interactome (DSI)” where commonly shared domains or motifs, such “SH3” and “PDZ domain” are two examples of frequently shared domains or motifs that are being screened
86
A. Alam et al.
and cloned for their interacting partners. The possibility of modifications to the physical and metabolic characteristics in a physiological context that changes protein binding at these domains is a limitation of this strategy. In addition, the majority protein–protein interactions (PPIs) within the current interactome are predicted based on the induced protein expression levels observed in experimental yeast cells, which could be very different from the endogenous ecosystem where key targets are normally expressed. An area of active research is the integration of PPIs with gene expression data and further techniques to provide tissue and disease-specific perspective to the existing interactome. Notably, the present interactome only contains one isoform of each gene product and provides scant annotations regarding the splice isoforms that are being considered. It is generally accepted that distinct spliced forms can results in significant changes in phenotypic variations. Reticulocyte analysis is a recently developed idea in which a patient’s unique integrative biological network (reticulome) and a set of molecular-mutants/variants are investigated. Each person has a unique set of biological networks. Biological network environment within an individual unquestionably influences the final result (phenotype) of a certain set of genetic variations (genotype) and therefore, should be an essential element of any patient-specific data assessment. Therefore, customized reticulotype-based network studies have potential to strengthen the current genotype/phenotype correlation attempts and may make the search for customized targeted therapies.
5.6 Case Studies 5.6.1 Case Study: 1 Imam et al. conducted a recent study titled “Network-Medicine Approach for the Identification of Genetic Association of Parathyroid Adenoma with Cardiovascular Disease and Type-2 Diabetes.” This study delves into unexplored dimensions of diseases by specifically investigating distantly related protein sets associated with other diseases that have not been previously studied. The aim is to understand their collective physiological impact on the pathological phenotype through network analysis. The researchers performed a comparative analysis of disease-associated proteins in Parathyroid Adenoma, Cardiovascular Disease, and Type-2 Diabetes with the aim of identifying shared genetic factors. Utilizing network analysis methods, they investigated functional modules within the protein-disease network that exhibited dense internal connections but sparse connections with the rest of the network. As a result, they discovered 13 target proteins that were found to be common to parathyroid adenoma, cardiovascular disease, and type 2 diabetes. These proteins were subsequently organized into hierarchical modules and sub-modules within the Protein–Protein Interaction (PPI) network. In their study, the researchers also employed the concept of drug repurposing and drug combinations. They utilized
5 Network Medicine: Methods and Applications
87
target proteins and associated drugs in a drug–target bipartite network, which included experimentally verified drug–target binary connections. They found that 36 drugs were common to both target-associated drugs (TAD) and disease-associated drugs (DAD), supporting the effectiveness of a multi-target drug approach. This network-based analysis presents promising avenues for personalized treatment and the repurposing of drugs. It allows for the exploration of new targets and combinations of multiple drugs, facilitating a comprehensive understanding of protein–disease associations and disease–disease relationships. By utilizing advanced computational techniques, the study prioritizes drug–target interactions and examines disease–disease connections, thereby enhancing the selection of potential therapeutic targets based on efficacy and safety in complex disease scenarios (For more details, read the full article: https://doi.org/10.1093/bfgp/elac054).
5.6.2 Case Study: 2 A study conducted by Aftab et al. aimed to construct a disease network by investigating the overlap between Tuberculosis (TB) and other Non-Communicable Diseases (NCDs) such as Parkinson’s Disease (PD), Cardiovascular Disease (CVD), Diabetes Mellitus (DM), Rheumatoid Arthritis (RA), and Lung Cancer (LC). Through the analysis of this disease network, the researchers identified common genes associated with TB and other NCDs, establishing important gene–disease relationships. To delve deeper into these diseases, the researchers constructed separate gene interaction networks for each disease by integrating carefully curated and experimentally validated human interactions. Additionally, they generated and analyzed a drug–target interactome network that encompassed clinically relevant drug–drug and drug–target interactions. This comprehensive network offered a more comprehensive understanding of the intricate landscape of drug–target interactions. The primary objective of this study was to establish a comprehensive workflow that takes into account Tuberculosis (TB) and its overlapping Non-Communicable Diseases (NCDs), emphasizing the significance of reconsidering and redefining therapies and therapeutic management. The findings underscore the potential of exploring uncharted territories in disease research, particularly the shared gene sets that coexist among various diseases. By fostering collaboration, it becomes feasible to collectively impact the pathological phenotype at a physiological level (For more details, read the full article: https://doi.org/10.3389/fphar.2021.770762).
5.7 Conclusion Multi-omics data will soon be regularly used in preclinical and clinical contexts, hence further constraints should apply standardized, rigorous bioinformatics techniques to process, normalize, and analyze the massive datasets from various study modalities. From this perspective, it should soon be possible to build networks of
88
A. Alam et al.
networks to determine how various biological aspects are interconnected. Beyond intracellular molecular networks, we are now combining the connections. between various cell-_types, organ-systems, hosts and microorganisms, hosts and environmental exposures, as well as other interconnections. It would also be essential to incorporate psychosocial components when defining disease networks in order to link all contributing factors to clinical results. As demonstrated, network controllability analysis continue to be useful for identifying important disease genes and prioritizing potential targets. Reticulocyte analysis is a recently developed idea in which a patient’s unique integrative biological network (reticulome) and a set of molecular-mutants/variants are investigated. Each person has a unique set of biological networks. A biological network environment within an individual unquestionably influences the final result (phenotype) of a certain set of genetic variations (genotype) and therefore, should be an essential element of any patient-specific data assessment. Therefore, customized reticulotype-based network studies have the potential to strengthen the current genotype–phenotype correlation attempts and may make the search for customized targeted therapies.
References Alam A et al (2019) Identification and classification of differentially expressed genes and network meta-analysis reveals potential molecular signatures associated with tuberculosis. Front Genet 10:932. https://doi.org/10.3389/fgene.2019.00932 Alam A, Imam N, Siddiqui MF, Ali MK, Ahmed MM, Ishrat R (2021) Human gene expression profiling identifies key therapeutic targets in tuberculosis infection: a systematic network meta- analysis. Infect Genet Evol 87:104649. https://doi.org/10.1016/j.meegid.2020.104649 Alam A et al (2022) An integrative network approach to identify common genes for the therapeutics in tuberculosis and its overlapping non-communicable diseases. Front Pharmacol 12:770762. https://doi.org/10.3389/fphar.2021.770762 Alanis-Lobato G, Andrade-Navarro MA, Schaefer MH (2017) HIPPIE v2.0: enhancing meaningfulness and reliability of protein–protein interaction networks. Nucleic Acids Res 45(D1):D408–D414. https://doi.org/10.1093/nar/gkw985 Alonso-López D et al (2019) APID database: redefining protein–protein interaction experimental evidences and binary interactomes. Database 2019:baz005. https://doi.org/10.1093/ database/baz005 Azuaje FJ, Zhang L, Devaux Y, Wagner DR (2011) Drug-target network in myocardial infarction reveals multiple side effects of unrelated drugs. Sci Rep 1(1):52. https://doi.org/10.1038/ srep00052 Bader GD (2003) BIND: the biomolecular interaction network database. Nucleic Acids Res 31(1):248–250. https://doi.org/10.1093/nar/gkg056 Bader GD, Hogue CW (2003) An automated method for finding molecular complexes in large protein interaction networks. BMC Bioinformatics 4(1):2. https://doi.org/10.1186/1471-2105-4-2 Black JW, Leff P (1983) Operational models of pharmacological agonism. Proc R Soc Lond B Biol Sci 220(1219):141–162. https://doi.org/10.1098/rspb.1983.0093 Boran ADW, Iyengar R (2010) Systems pharmacology. Mt Sinai J Med 77(4):333–344. https://doi. org/10.1002/msj.20191 Calderone A, Castagnoli L, Cesareni G (2013) Mentha: a resource for browsing integrated protein- interaction networks. Nat Methods 10(8):690–691. https://doi.org/10.1038/nmeth.2561 Chatr-aryamontri A et al (2007) MINT: the molecular interaction database. Nucleic Acids Res 35, no. Database:D572–D574. https://doi.org/10.1093/nar/gkl950
5 Network Medicine: Methods and Applications
89
Colquhoun D (2006) The quantitative analysis of drug-receptor interactions: a short history. Trends Pharmacol Sci 27(3):149–157. https://doi.org/10.1016/j.tips.2006.01.008 Costa AJ (1991) Potential drug interactions in an ambulatory geriatric population. Fam Pract 8(3):234–236. https://doi.org/10.1093/fampra/8.3.234 Dara S, Dhamercherla S, Jadav SS, Babu CM, Ahsan MJ (2022) Machine learning in drug discovery: a review. Artif Intell Rev 55(3):1947–1999. https://doi.org/10.1007/s10462-021-10058-4 Das J, Yu H (2012) HINT: high-quality protein interactomes and their applications in understanding human disease. BMC Syst Biol 6(1):92. https://doi.org/10.1186/1752-0509-6-92 Fahey ME et al (2011) GPS-Prot: a web-based visualization platform for integrating host-pathogen interaction data. BMC Bioinformatics 12(1):298. https://doi.org/10.1186/1471-2105-12-298 Galletti C, Bota PM, Oliva B, Fernandez-Fuentes N (2021) Mining drug-target and drug-adverse drug reaction databases to identify target-adverse drug reaction relationships. Database (Oxford) 2021:baab068. https://doi.org/10.1093/database/baab068 Ghiassian SD, Menche J, Barabási A-L (2015) A disease module detection (DIAMOnD) algorithm derived from a systematic analysis of connectivity patterns of disease proteins in the human interactome. PLoS Comput Biol 11(4):e1004120. https://doi.org/10.1371/journal. pcbi.1004120 Goel R, Harsha HC, Pandey A, Prasad TSK (2012) Human protein reference database and human proteinpedia as resources for phosphoproteome analysis. Mol Bio Syst 8(2):453–463. https:// doi.org/10.1039/C1MB05340J Hermjakob H (2004) IntAct: an open source molecular interaction database. Nucleic Acids Res 32(90001):452D–D455. https://doi.org/10.1093/nar/gkh052 Imam N, Alam A, Siddiqui MF, Veg A, Bay S, Khan MJI, Ishrat R (2023) Network-medicine approach for the identification of genetic association of parathyroid adenoma with cardiovascular disease and Type-2 diabetes. Brief Funct Genomics 22(3):250–262. https://doi.org/10.1093/ bfgp/elac054 Kalathur RKR et al (2014) UniHI 7: an enhanced database for retrieval and interactive analysis of human molecular interaction networks. Nucl Acids Res 42(D1):D408–D414. https://doi. org/10.1093/nar/gkt1100 Kim K-J, Moon S-J, Park K-S, Tagkopoulos I (2020) Network-based modeling of drug effects on disease module in systemic sclerosis. Sci Rep 10(1):13393. https://doi.org/10.1038/ s41598-020-70280-y Kotlyar M, Pastrello C, Sheahan N, Jurisica I (2016) Integrated interactions database: tissue-specific view of the human and model organism interactomes. Nucleic Acids Res 44(D1):D536–D541. https://doi.org/10.1093/nar/gkv1115 Lee LY-H, Loscalzo J (2019) Network medicine in pathobiology. Am J Pathol 189(7):1311–1326. https://doi.org/10.1016/j.ajpath.2019.03.009 Luck K et al (2020) A reference map of the human binary protein interactome. Nature 580(7803):402–408. https://doi.org/10.1038/s41586-020-2188-x Maehle A-H, Prüll C-R, Halliwell RF (2002) The emergence of the drug receptor theory. Nat Rev Drug Discov 1(8):637–641. https://doi.org/10.1038/nrd875 Morselli Gysi D et al (2021) Network medicine framework for identifying drug-repurposing opportunities for COVID-19. Proc Natl Acad Sci USA 118(19):e2025581118. https://doi. org/10.1073/pnas.2025581118 Newman MEJ (2006) Finding community structure in networks using the eigenvectors of matrices. Phys Rev E 74(3):036104. https://doi.org/10.1103/PhysRevE.74.036104 Oughtred R et al (2019) The biogrid interaction database: 2019 update. Nucleic Acids Res 47(D1):D529–D541. https://doi.org/10.1093/nar/gky1079 Patil A, Nakai K, Nakamura H (2011) HitPredict: a database of quality assessed protein–protein interactions in nine species. Nucleic Acids Res 39(suppl_1):D744–D749. https://doi. org/10.1093/nar/gkq897 Pinero J et al (2015) DisGeNET: a discovery platform for the dynamical exploration of human diseases and their genes. Database 2015(0):bav028. https://doi.org/10.1093/database/bav028
90
A. Alam et al.
Sakle NS, More SA, Mokale SN (2020) A network pharmacology-based approach to explore potential targets of Caesalpinia pulcherima: an updated prototype in drug discovery. Sci Rep 10(1):17217. https://doi.org/10.1038/s41598-020-74251-1 Schoof H et al (2005) Munich information center for protein sequences plant genome resources. a framework for integrative and comparative analyses. Plant Physiol 138(3):1301–1309. https:// doi.org/10.1104/pp.104.059188 Szklarczyk D et al (2019) STRING v11: protein–protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res 47(D1):D607–D613. https://doi.org/10.1093/nar/gky1131 Vella D, Marini S, Vitali F, Di Silvestre D, Mauri G, Bellazzi R (2018) MTGO: PPI network analysis via topological and functional module identification. Sci Rep 8(1):5499. https://doi. org/10.1038/s41598-018-23672-0 Wang Y, Qian X (2014) Functional module identification in protein interaction networks by interaction patterns. Bioinformatics 30(1):81–93. https://doi.org/10.1093/bioinformatics/btt569 Xenarios I (2000) DIP: the database of interacting proteins. Nucleic Acids Res 28(1):289–291. https://doi.org/10.1093/nar/28.1.289 Xiong G et al (2022) DDInter: an online drug–drug interaction database towards improving clinical decision-making and patient safety. Nucleic Acids Res 50(D1):D1200–D1207. https://doi. org/10.1093/nar/gkab880
6
Role of R in Biological Network Analysis Mohd Murshad Ahmed and Safia Tazyeen
Abstract
The life sciences are becoming increasingly data rich as a result of technical development in molecular biology (large-scale systems biology, genomics). Appropriate analysis of large-scale datasets (proteomics, genomes, multi-omics data, Htseq, RNA-seq, ChIPseq, and so on) is now a bottleneck. One of the most frequent ways of representing biological systems as complicated sets of binary interactions or relationships between diverse bio-entities is through networks. In fields like computational biology, finance, neurology, political science, and public health, network analysis is a technique that employs graph theory to investigate complicated real-world hurdles. Networks, which are mainly portrayed as graphs with hundreds of nodes and thousands of vertices, are the best way to describe the various components of a system and their interactions. Some of the potential applications of network analysis in biology and medicine include discovering drug targets, determining the function of a protein or gene, inventing successful strategies for treating various diseases, and providing early identification of abnormalities. Several software tools, such as Gephi and Cytoscape, are built for network analysis and the production of network graphs. R has evolved into a formidable tool for network analysis, despite not being specifically designed for it. R has three distinct advantages over standalone network analysis applications. The goal of this research/chapter is to provide a solid understanding of how to use R to perform advanced data analysis. Researchers will learn how to extract relevant information from multiple datasets using a variety of regularly
M. M. Ahmed (*) Singapore Institute of Clinical Sciences (SICS), Singapore, Singapore S. Tazyeen Center for Interdisciplinary Research in Basic Sciences, Jamia Millia Islamia, New Delhi, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 R. Ishrat (ed.), Biological Networks in Human Health and Disease, https://doi.org/10.1007/978-981-99-4242-8_6
91
92
M. M. Ahmed and S. Tazyeen
used statistical and mathematical methodologies. The goal is also to demonstrate and highlight a few important methodologies related to network analysis, rather than to provide a fully-fledged analysis. In this chapter, we illustrate ideas, models, and methods from the graph theory world and explore how they might be utilized to uncover hidden aspects and features of a network. Keywords
Packages · Co-expression network · Survival analysis · Modules analysis · Gene ontology
6.1 Introduction In biology, network structures are everywhere because many biological systems depend on intricate interactions between the parts that make them up. Predator–prey partnerships are the most frequent kind of species interaction in ecosystems and are essential to preserving biodiversity (Oleskin 2014). Synapses in the human brain are used to transmit electrical and chemical signals between neurons. DNA, RNA, proteins, and other components play various roles in biochemical processes at the cellular level that control how cells function. These systems are best represented mathematically by networks, which have sets of elements, which we shall call vertices or nodes, with connections between them, called edges (Wang et al. 2021). Discovering the biological information from network concepts is of major significance since biological entities are involved in complicated and complex interactions (Liu et al. 2020). Experimental biology developments have increased the availability of large-scale biological network data (Vitkup 2004). We examine networks of protein–protein interactions (PPIs), in which proteins are represented as network nodes and interactions as network edges (Nabieva et al. 2005). Since proteins interact with one another to carry out nearly all biological functions, studying the structure of PPI networks may provide new information about disease and complex biological phenomena (Memisevic et al. 2010). Researchers have gained fresh insights into the challenge of modeling complex systems as a result of the growth of network data in a variety of fields, which has also sparked the development of various cutting-edge statistical approaches and computational tools/databases such as R/RStudio/GEO2R (Wang et al. 2021). R is a widely used open-source programming language for statistical computing and data analysis. R typically includes a command-line interface (Lovelace et al. 2019). R is accessible on popular operating systems like Windows, Linux, and macOS. The newest cutting-edge technology is the R programming language developed by Ross Ihaka and Robert Gentleman (Giorgi et al. 2022; Ihaka and Gentleman 1996). Much of the intricacy of cellular life is mediated by gene interaction networks, in which genes activate and repress the transcription of other genes, and their dysfunction can be fatal to an organism. So, systems biology has long sought to understand these networks (Saint-Antoine and Singh 2020). Exploring gene interactions experimentally by testing every possible gene pair is impractical due to the large number of genes present in the human
6 Role of R in Biological Network Analysis
93
genome, which is approximately 20,000. However, modern molecular biology techniques, such as DNA microarrays and next-generation sequencing (NGS), have made it possible to obtain a quantitative understanding of the transcriptome profile of a single cell or a group of cells. RNA-sequencing (RNA-Seq), whether done on individual cells or as bulk sequencing, is a particularly valuable tool for this purpose. These methods have greatly facilitated the discovery of gene interactions and other molecular mechanisms underlying complex biological processes (Madsen et al. 2022). This chapter will demonstrate the numerous applications of the R language. Despite the fact that R is one of the most popular and effective programming languages in bioinformatics, R excels in the creation of graphs and figures of publication quality and in the use of a variety of statistical tools, such as RNA-Seq and genomics.
6.2 Installation of R Software and Packages We will start by learning the statistical programming language R. R programming is a great option for processing and analyzing life science data since it has a large community of developers who are constantly creating new R programming packages. R software is deserving widespread acclaim, and it will continue to grow. A wide range of statistical techniques is supported by R software, including traditional statistical tests, modeling (both linear and nonlinear), classification, time series analysis, cluster analysis, and graphical data display. R software is a great option for manipulating big data and life science data because it is extremely flexible and simple to learn. We first need to get access to R, then we can begin learning it. The focus of this chapter will be on R implementation. An R installation is available at http://cran.r-project.org and RStudio at http://www.rstudio.com/. This is simple; the software package is not large, and I normally work on things offline. R console like below will appear, and the R program will get started, see Fig. 6.1.
6.2.1 Packages in R R packages are a collection of preset functions that can be used as a library when deploying an R application to promote reuse and a low-code design. R packages are created outside of the R environment and can be imported to make use of the available functions they include. When a package is loaded into the R environment, it offers the necessary functions that can be used. A library in C, C++, or Java is analogous to a list of R Packages. Therefore, a package might essentially comprise a variety of functionality, such as functions, and constants, that we will then let the user use in the context of a certain situation. R comes with a number of packages that may be obtained from CRAN (Comprehensive R Archive Network) and GitHub.
Fig. 6.1 The diagram depicts the information about R. Sixty-four bit downloaded and installed in the window PC
94 M. M. Ahmed and S. Tazyeen
6 Role of R in Biological Network Analysis
95
6.2.2 Installing R Packages Direct package installation is possible using either the IDE or commands. We use the function below and the package name to install packages: Syntax: install.packages()
Code: install.packages(“affy”)
The above code installs the affy package in R.
6.2.3 Loading R Packages To use the loaded packages in R, we must load them after installing the R package. To load the packages, we employ the functions below: Syntax: >Library (package name)
Code: >Library (affy)
Now affy package is loaded in R Console. To overview the complete process of R installation and update refer to Fig. 6.2.
6.3 To Build Network Model Using R Data must be in a specific form for the network analysis packages to build the unique type of object that is required by each program. Adjacency matrices, also known as sociomatrices, are the foundation for the object classes used by the programs network, igraph, and tidygraph (see Fig. 6.3). A three-edged, undirected graph is produced by the code below. The edges are 1- > 2, 2- > 3, 3- > 1 since the integers are read as vertex IDs. M1